Grab bag of threading fixes:

1. Remove the coop two-phase transition (Partially revert 6726fae1).  This was based on a misunderstanding of how Emscripten works: when the main thread is blocked in a concurrency primitive like `sem_wait`, it is still processing queued calls from other threads.  So there is no need to first suspend the worker threads and then suspend the main thread.  The implementation of two-phase suspend had a bug where it would suspend worker threads twice, making the suspend increase by 2. Since resume only decremented the count by 1, this lead to a suspend count overflow.  Fixes #72857 
2. Once the diagnostic server attaches to the runtime, switch it to GC Safe mode when it returns to JavaScript.  That is, while the diagnostic server is reacting to messages in the JS event loop, it is considered suspended by the runtime.  When it calls into C, switch to GC Unsafe (which may block if there's a STW happening).  Add thread state transitions when we come back to C, and when we wait.
3. Mark the wasm diagnostic server thread as "no sample; no gc" which means that we don't consider it for STW when there's a GC or a sample profiler active.  This is how we treat utility threads (including the non-wasm diagnostic server thread) on other platforms.
4. Fix a bad signature for `cwraps.mono_wasm_event_pipe_enable` due to a mistake in a previous merge
5. Added a new `browser-threads` sample

---

* [coop] Don't call begin_suspend_request_suspension_cordially twice

Each call to begin_suspend_request_suspension_cordially may increment
the suspend count.  But in STW we only resume each thread once.

In 6726fae1 we added a second phase of
STW to full coop on WebAssembly in order to suspend the browser thread
after all the worker threads have been suspended in order to avoid
some deadlocks that rely on the main thread continuing to process
async work on behalf of the workers before they reach a safepoint.

The problem is that for worker threads we could end up calling
begin_suspend_request_suspension_cordially twice. If the thread
self-suspends after the first call, the second call will increment the
suspend count.  As a result, when we restart the world, the thread
will decrement its suspend count, but still stay suspended.  Worse, on
the _next_ stw, we will increment the suspend count two more times and
decrement once on the next restart, etc.

Eventually the thread will overflow the suspend counter and we will
assert `!(suspend_count > 0)`.

Also change `THREAD_SUSPEND_COUNT_MAX` to `0x7F` (from `0xFF`) - the
suspend count is signed, so the roll-over from 127 to -128 is where we
should assert

Fixes https://github.com/dotnet/runtime/issues/72857

* improve thread state machine assertion messages

   include thread states and thread ids where available

* fix typo

* Revert "[coop] Don't call begin_suspend_request_suspension_cordially twice"

This reverts commit 92f52ab7ed1cfaa1a4f66e869a8d9404e066f1b2.

* [threads] Revert coop two-phase STW

Remove mono_threads_platform_stw_defer_initial_suspend

The motivation for it in 6726fae1
was unfounded.  There is no need to suspend the main browser thread
after the other threads: suspension on wasm uses `sem_wait` which on
Emscripten on the main thread is implemented using a busy wait
`__timedwait_cp` which processes queued calls.  So even if we suspend
the main thread first, it will still allow other threads in GC Safe to
make progress if they're using syscalls.

* Switch the diagnostics server to GC Safe when returning to JS; set NO_GC flag

The diagnostic server worker spends most of its time in the JS event
loop waiting for messages.  After we attach to the runtime, we need to
switch to GC Safe mode because the diagnostic server may not ever
reach a safepoint (for example if no more DS events arrive).

Conversely, when we call from JS into the C diagnostic server, we need
to enter GC Unsafe mode (and potentially safepoint).

Also mark the diagnostic server threads with the NO_GC flag - this
thread does not manipulate managed objects so it doesn't need to stop
for GC STW.

* cwraps: fix bad signature for mono_wasm_event_pipe_enable

   Mistake from a previous merge

* Add new browser-threads sample

* exclude the browser-threads sample unless wasm threads are enabled

* Update browser-threads sample

* Update src/mono/mono/component/diagnostics_server.c

* Update src/mono/mono/utils/mono-threads-state-machine.c
Co-authored-by: NKatelyn Gadd <kg@luminance.org>

Fixes https://github.com/dotnet/runtime/issues/50365

This reverts commit 73cf30dd.

As a result of #73013 landing, the skipped StructureToPtr tests can be enabled.

Contributes to #73008

SendMailAsync_CanBeCanceled_CancellationToken is failing also on CoreCLR and on OSX - tracked by #73447

Fix #61411 and harmonize argument exception handling.

* Fix sending end stream for websocket over H/2

* feedback

* Update src/libraries/System.Net.WebSockets.Client/tests/SendReceiveTest.Http2.cs
Co-authored-by: NNatalia Kondratyeva <knatalia@microsoft.com>

* default usessl in tests

* Update src/libraries/System.Net.WebSockets.Client/tests/ConnectTest.Http2.cs
Co-authored-by: NStephen Toub <stoub@microsoft.com>

* disable new tests on browser because sockets not supported
Co-authored-by: NNatalia Kondratyeva <knatalia@microsoft.com>
Co-authored-by: NStephen Toub <stoub@microsoft.com>

Fix #73409

* Fix global.json empty parse error by adding brackets
* Update the version of dotnet being used to build Maui apps and the rollback file to a set of compatible workload versions.

* Vectorize Convert.ToBase64CharArray and TryToBase64Chars

A previous PR vectorized Convert.ToBase64String for larger inputs by using Base64.EncodeToUTF8 and then encoding the result UTF8 bytes into a UTF16 string.  It did not touch Convert.ToBase64CharArray nor Convert.TryToBase64Chars, however.  The ToBase64String change makes use of a temporary array rented from the array pool, and the expectation is it'll rarely allocate, but if it does, it's part of a method that's already allocating the resulting string and so it's presumed to not be too impactful.  ToBase64CharArray and TryToBase64Chars, however, are intended to be entirely non-allocating, and so even renting from the array pool would be problematic.

This PR changes the non-allocating variants to use Base64.EncodeToUtf8 as well.  But instead of renting a temporary buffer, it banks on the knowledge that the encoded Base64 bytes are 1/2 the length of the resulting chars, since the bytes are all guaranteed to be ASCII.  Thus, it can treat the destination char buffer as scratch space for the encoded UTF8 bytes, and then widen in-place.  This obviates the need for a separate temporary buffer, making it appropriate for the non-allocating versions.  And once we had the helper for those, we can use that same helper to replace the code added to ToBase64String, making it non-allocating as well (beyond of course the result string it has to allocate by its nature), and thus making it more predictable.

Overall, this fixes the possible additional allocation in ToBase64String as well as the performance inversion that the allocating ToBase64String could have been significantly faster (due to vectorization) than the ToBase64CharArray and ToBase64Chars methods intended to be the faster versions.

* Address PR feedback and a bit of additional cleanup

- Disable test that is testing 64-bit incompatible API on 64-bit platforms
- Relax ArgumentException argument name validation. The name of the argument that test was expecting did not match the actual name of the method argument since the ArgumentException originates several layers deep into the implementation.

Fixes https://github.com/dotnet/runtime/issues/73421
Fixes https://github.com/dotnet/runtime/issues/73422
Fixes https://github.com/mono/mono/issues/15311

* Disambiguate calli thunks based on full calling conventions

Fixes dotnet/runtimelab#153

* Add smoke test
Co-authored-by: NMichal Strehovský <MichalStrehovsky@users.noreply.github.com>

* Documenting providing websocket upgrade response

* Apply suggestions from code review
Co-authored-by: NMiha Zupan <mihazupan.zupan1@gmail.com>
Co-authored-by: NMiha Zupan <mihazupan.zupan1@gmail.com>

Disable System.Net.Quic.Tests.QuicStreamConformanceTests.Parallel_ReadWriteMultipleStreamsConcurrently (#73446)

.. builds.

https://github.com/dotnet/runtime/pull/72394 disabled the cache always,
even when it was intended to be used. This reverses that change.

Fixes https://github.com/dotnet/runtime/issues/73419

* Implement thread-local caching for Utf8JsonWriter in JsonSerializer.

* address feedback

* Address feedback

* Fix nullability warnings in ns2.0 build

The sdk issue https://github.com/dotnet/sdk/issues/26967 was fixed in
https://github.com/dotnet/sdk/pull/26796 .

Fixes https://github.com/dotnet/runtime/issues/73312 .

This time found some product issues and fixed them.

- This support has been a bit of a bug farm, and hasn't provided much value as we have the more capable Mibc support
- Customers which wish to use Ibc support can now enable this support as needed

* Adding the IBinaryInteger<T>.ReadLittleEndian and IBinaryInteger<T>.ReadBigEndian APIs

* Adding tests for the IBinaryInteger<T>.ReadLittleEndian and IBinaryInteger<T>.ReadBigEndian APIs

* Ensure IBinaryInteger<T>.ReadBig/LittleEndian APIs take ROSpan

* Update src/libraries/System.Private.CoreLib/src/System/Byte.cs
Co-authored-by: NStephen Toub <stoub@microsoft.com>

* Fix accidentally removed 0 in char

* Update IndexOfAny to compare against >= 0, rather than != -1
Co-authored-by: NStephen Toub <stoub@microsoft.com>

As Andy Ayers suggested in the issue #72822 and I confirmed,
Crossgen2 signature emitter was missing the code path for
emitting function pointer signatures. This change fixes this
deficiency. The change also removes the issues.targets exclusion
of the affected test that Andy originally merged in to mitigate
the issue.

Thanks

Tomas

* increase timeout on ZeroByteRead_IssuesZeroByteReadOnUnderlyingStream test'

* feedback from review

* feedback from review

* Update InteropServices tests for NativeAOT

* Remove SetWin32ContextInIDispatchAttribute attribute and tests.
  - this was no longer in the ref assembly so just removing it.

* Remove dependency on registered COM server. This enables more
  expansive testing on all Windows SKUs.

* Fail JsonTypeInfo.CreateObject setting when not supported by converter.

* add clarifying comment.

* Remove dead code.

* Fix InvalidCastException when casting a polymorphic converter.

* Extend test coverage and fixes for the case of deserialization.

* Update src/libraries/System.Text.Json/src/System/Text/Json/Serialization/Metadata/JsonTypeInfoOfT.cs

Add TieredPGO and ReadyToRun to Microsoft-Windows-DotNETRuntime/TieredCompilation/Settings event (#72056)