Misc changes related to Miri allocations

This builds on top of https://github.com/rust-lang/rust/pull/50249

r? @oli-obk

Filter global bounds from ParamEnv again.

This PR adds back the filtering of global bounds from ParamEnv as a temporary solution for #50825.

<details>

Long term, the fix seems like it should be changing the priority in `candidate_should_be_dropped_in_favor_of` so that (global) where clauses aren't considered as highly.

https://github.com/rust-lang/rust/blob/a722296b6ec17fecd3f16a7d3f9232b83e5de800/src/librustc/traits/select.rs#L2017-L2022

</details>

r? @nikomatsakis

rustc: Fix joint-ness of stringified token-streams

This commit fixes `StringReader`'s parsing of tokens which have been stringified
through procedural macros. Whether or not a token tree is joint is defined by
span information, but when working with procedural macros these spans are often
dummy and/or overridden which means that they end up considering all operators
joint if they can!

The fix here is to track the raw source span as opposed to the overridden span.
With this information we can more accurately classify `Punct` structs as either
joint or not.

Closes #50700

Stabilize suggestion applicability field in json output

cc @killercup

r? @nrc

Escape combining characters in char::Debug

Although combining characters are technically printable, they make little sense to print on their own with `Debug`: it'd be better to escape them like non-printable characters.

This is a breaking change, but I imagine the fact `escape_debug` is rare and almost certainly primarily used for debugging that this is an acceptable change.
Resolves #41922.

r? @alexcrichton
cc @clarcharr

Fix issue #50811 (`NaN > NaN` was true).

Fix #50811

Make sure the float comparison output is consistent with the expected behavior when NaN is involved.

----

Note: This PR is a **BREAKING CHANGE**. If you have used `>` or `>=` to compare floats, and make the result as the length of a fixed array type, like:

```rust
use std::f64::NAN;
let x: [u8; (NAN > NAN) as usize] = [1];
```

then the code will no longer compile. Previously, all float comparison involving NaN will just return "Greater", i.e. `NAN > NAN` would wrongly return `true` during const evaluation. If you need to retain the old behavior (why), you may replace `a > b` with `a != a || b != b || a > b`.

Quick refactoring around Substs & friends.

r? @nikomatsakis

stabilize opt-level={s,z}

closes #35784
closes #47651

### Rationale

Since the lastest LLVM upgrade rustc / LLVM does more agressive loop unrolling. This results in increased binary size of embedded / no_std programs: a hundreds of bytes increase, or about a 7x increase, in the case of the smallest Cortex-M binary cf. #49260.

As we are shooting for embedded Rust on stable it would be great to also provide a way to optimize for size (which is pretty important for embedded applications that target resource constrained devices) on stable.

Also this has been baking in nightly for a long time.

r? @alexcrichton which team has to sign off this?

UnsafeCell doc typos and minor flow improvements

r? @QuietMisdreavus

Switch Vec from doubling size on growth to using RawVec's reserve

On growth, Vec does not require to exactly double its size for correctness,
like, for example, VecDeque does.

Using reserve instead better expresses this intent. It also allows to reuse
Excess capacity on growth and for better growth-policies to be provided by
RawVec.

r? @sfackler

lexer: Fix span override for the first token in a string

Previously due to peculiarities of `StringReader` construction something like `"a b c d".parse::<TokenStream>()` gave you one non-overridden span for `a` and then three correctly overridden spans for `b`, `c` and `d`.
Now all the spans are overridden.

Find the largest niche when computing layouts

Otherwise we end up with `Option<Option<(&(), bool)>>` unnecessarily large.

rustc: introduce {ast,hir}::AnonConst to consolidate so-called "embedded constants".

Previously, constants in array lengths and enum variant discriminants were "merely an expression", and had no separate ID for, e.g. type-checking or const-eval, instead reusing the expression's.

That complicated code working with bodies, because such constants were the only special case where the "owner" of the body wasn't the HIR parent, but rather the same node as the body itself.
Also, if the body happened to be a closure, we had no way to allocate a `DefId` for both the constant *and* the closure, leading to *several* bugs (mostly ICEs where type errors were expected).

This PR rectifies the situation by adding another (`{ast,hir}::AnonConst`) node around every such constant. Also, const generics are expected to rely on the new `AnonConst` nodes, as well (cc @varkor).
* fixes #48838
* fixes #50600
* fixes #50688
* fixes #50689
* obsoletes #50623

r? @nikomatsakis