Allow the inliner to substitute for small arguments

[jakobbotsch]

Since we no longer have contextual nodes to indicate small arguments
this should be safe to do. For register arguments it does not matter and
for stack arguments it is handled in codegen (for macOS arm64).

[ghost]

Tagging subscribers to this area: @JulieLeeMSFT
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Since we no longer have contextual nodes to indicate small arguments
this should be safe to do. For register arguments it does not matter and
for stack arguments it is handled in codegen (for macOS arm64).

Author:	jakobbotsch
Assignees:	jakobbotsch
Labels:	area-CodeGen-coreclr
Milestone:	-

[jakobbotsch]

/azp run Fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jakobbotsch]

This is waiting on #69117

[jakobbotsch]

/azp run Fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jakobbotsch]

/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[jakobbotsch]

Failures are #69349, #68690, #69350, #69232, #69154, #68513

[jakobbotsch]

cc @dotnet/jit-contrib PTAL @AndyAyersMS
I need to spot check the diffs, but this is generally an improvement and just undoes the limitation added in #43130.

[jakobbotsch]

Spot-checked some of the regressions, they fell in the following categories:

1. Different (worse) register allocation, with some additional spills
2. Different alignment
3. We create more locals now (that we later may substitute if they are single-used) -- more locals lead to different inlining decisions (Inliner uses capacity of locals table instead of count #69280 related)
4. Fewer statements in BBs mean we sometimes duplicate more conditional jumps (fgOptimizeUncondBranchToSimpleCond) -- this leads to more code but a shorter critical path