JIT: Fix tailcall-to-loop improper locals zeroing

[jakobbotsch]

The zeroing that the tailcall-to-loop optimization does was zeroing the
promoted copies for implicit byrefs even when promotion of them was undone.
This was introducing unexpected references to the promoted fields.

Fix#81081

[ghost]

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

Issue Details

---

The zeroing that the tailcall-to-loop optimization does was zeroing the
promoted copies implicit byrefs even when promotion of them was undone.
This was introducing unexpected references to the promoted fields.

Fix #81081

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

[jakobbotsch]

The arm32 tpdiff was caused by this error:

[16:21:44] ERROR:Method423038 of size 7557059 failed to load and compile correctly by JIT2 (C:\h\w\A764090A\p\diff\release\clrjit_universal_arm_x86.dll).
[16:21:44] ERROR: main method 423038 of size 7557059 failed to load and compile correctly.
[16:21:44] ERROR:Exceptionthrown: SuperPMI assertion 'index!=-1' failed ("Didn't find Item (in Get)")[16:21:44] failed (missing key "key" in map GetObjectType): key FFFFFFFFEE627D78

I cannot reproduce it locally, this was probably some OOM error (the exception code looks like a C++ exception and the size looks very large).
There is a separate issue which is that SPMI does not look to properly handle discarding the diffs when the base JIT succeeds and then the diff JIT fails, I will add it on my todo list to fix.

The test failure looks like #76905.

[jakobbotsch]

cc @dotnet/jit-contrib PTAL @AndyAyersMS