Arm64-SVE: Fixes for returning Vector<T> through the return buffer

[snickolls-arm]

Invalidates debug information for return buffer when it holds TYP_SIMD. Computing an accurate stack offset is not possible in AOT, so we avoid introducing this support. Better information can be provided when the return value is passed in a vector register.

Prevent the hidden return buffer optimization when the local has TYP_SIMD. This optimization will no longer apply when the full vector calling convention is supported, so we avoid building support now.

[azure-pipelines]

Azure Pipelines:
Successfully started running 5 pipeline(s).
11 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

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

[snickolls-arm]

@dotnet/arm64-contrib Please could I have a review for this patch?

[jakobbotsch]

Prevent the hidden return buffer optimization when the local has TYP_SIMD. This optimization will no longer apply when the full vector calling convention is supported, so we avoid building support now.

What support is that? It is not clear to me if this isn't just hiding some bug that should be addresses regardless.

[jakobbotsch]

Is it just about the defSize part right below? If so then this seems ok.

[snickolls-arm]

Is it just about the defSize part right below? If so then this seems ok.

Yes, the assertion in GetSize() will fire without this change. I didn't want to update that to handle the possibility of inexact sizes, because we'd prefer to return through register z0 rather than the return buffer in future.