ARM64-SVE: Use predicate versions of instructions

[a74nh]

Some instructions provide versions which work on predicates. Eg, the final variant here:

BIC Ztied1.B, Pg/M, Ztied1.B, Zop2.B
BIC Zresult.D, Zop1.D, Zop2.D
BIC Presult.B, Pg/Z, Pop1.B, Pop2.B

Currently, when passing in a mask vector, coreclr will convert the mask to a normal vector and use the vector variant of the instruction.
Instead, coreclr should spot it's a mask and use the predicate variant.

If we don't do this, then instead we should remove the predicate variants from the summaries in Sve.cs

[dotnet-policy-service]

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

[kunalspathak]

@a74nh - Is #101598 one of the example of this issue?

[a74nh]

@a74nh - Is #101598 one of the example of this issue?

Yes, it's the same issue.
Not sure if you've a preference on which to keep open.

[a74nh]

priority:3 for RC1 snap : Performance issue, not functionality. Ideally we'd update the API summary if not doing this yet, but it's not the end of the world. I'd rather spend the time on something useful rather than something that will get reverted when this is properly fixed.

[a74nh]

This issue is dependant on #108241 being done first