Fix & test for #18362

[CarolEidt]

Also add to the arm & arm64 tests.lst

[CarolEidt]

The first commit is the test; it fails as expected on arm (it doesn't fail with the altjit).

[CarolEidt]

The problem was that LSRA was creating only a singe use for multi-reg ops under FIELD_LIST. We need to build a use for each reg.
Also, dump the defList if it's not empty at end of block, to make it easier to track down the cause of this assert in future.

[CarolEidt]

@dotnet/jit-contrib PTAL

[BruceForstall]

Link: #18362

[CarolEidt]

Fix #18362

[CarolEidt]

@BruceForstall @jashook the reason this didn't repro with the altjit is that getTailCallCopyArgsThunk returns null, so it doesn't do the tail copy (which leads to the split args case that this bug needs to repro). Do you think it's worth making it return a dummy value for the altjit?

[jashook]

I will let Bruce comment further; however, adding in altjit specific return values to me sounds dangerous.

[BruceForstall]

@CarolEidt I think we should make it work for altjit as you suggest. Wrap getTailCallCopyArgsThunk as eeGetTailCallCopyArgsThunk in ee_il_dll.cpp, then #ifdef on ARM and AMD64 under !info.compMatchedVM to return something dummy, to match the VM CEEInfo::getTailCallCopyArgsThunk conditions.