* MethodSymbol and FieldSymbol equality fixes:
 - Defer the equality of method/field symbol to the substituted symbol when compared
 - Make the hashcode of substituted method/field defer to the original definition

Rename from IncludeNonNullableReferenceTypeModifier to IncludeNotNullableReferenceTypeModifier (#41332)

Closes #41304

This API has always accepted null as a shortcut for an empty array, so
this annotates it accordingly.

Had to clean up a few nullable annotations now that we are compiling
agaist `netcoreapp3.1` and hence get the full value of the framework
annotations.

This is also problematic though because there are now two places where
the compiler can see nullable attributes that are directly used by the
developer. For example `NotNullWhenAttribute`. This is both defined in
our assemblies for non-netcoreapp target frameworks and provided by the
SDK when targeting `netcoreapp3.1`.

This causes a problem for assemblies which have the following
characteristics:

1. Target `netcoreapp3.1`
1. Reference an assembly targeting `netstandard2.0` which uses our
nullable attributes definition
1. Has IVT into (2) above

These properties essentially define all of our unit test assemblies. In
that environment it's not possible to use nullable attributes in code
because the compiler can't disambiguate which definition of
`NotNullWhenAttribute` to use. This meant I had to temporarily remove a
few attributes until we can complete #40766.

The TPA now considers the output directory as a trusted place for
loading assemblies. To the point it will prefer certain DLLs in the
output directory over equal ones that ship with the runtime.

This broke a number of VB tests here because we unconditionally
referenced the desktop version of Microsoft.VisualBasic. This got copied
to the output directory, included in the TPA and hence loaded during
test execution. This breaks tests because they require the .NET Core
version of Microsoft.VisualBasic.

Conditioned the reference to be desktop only so it's not present on our
.NET Core builds

Decided to remove the property based approach to specifying a target
framework to just specifying `netcoreapp3.1` directly. The reason for
this is the following:

The advantage of the property is it makes it "easy" to change to a new
target framework in the future. That benefit is actually pretty minimal.
A simple find and replace operation is **extremely** effective in our
code base (it's less key strokes than this message). Hence the benefit
is minimal.

The downside of the property is that our code doesn't look like customer
code. Or rather it diverges from the practices that we publish. In
general I prefer to keep our code as standard as possible unless there
is a good reason to deviate. There just doesn't seem to be one here.

https://github.com/microsoft/msbuild/issues/4806

Change Roslyn to target netcoreapp3.1 when building .NET Core assets.

Previously the code targetted a mix of netcoreapp2.1 and netcoreapp3.0.
The mix is due to default interfaces only being supported on
netcoreapp3.0 and hence our testing needed to use that. Yet at the same
time we were required to ship the compiler in SDKS that targetted
netcoreapp2.1. Now we can universally target netcoreapp3.1.

Fixes #40629

See #41273

* Move "is telemetry allowed" flag to DiagnosticAnalyzerInfoCache

* Clean up DiagnosticService

* Fixes

* Remove analyzer exception telemetry reporting.

* Remove reporting analyzer exceptions from SupportedDiagnostics.

The diagnostics will be reported when the compiler runs the analyzer on the compilation.

* Remove unused parameter

* Clean up analyzer telemetry reporting.

* Test fixes

* Feedback

Fixes #41207.