[llvm] [AArch64] Add C1-Premium scheduling model (PR #185398)

[Asher Dobrescu via llvm-commits]

https://github.com/Asher8118 commented:

Thank you for your changes, I've gone through all the comments and there are still a few that need addressing. Other than what has been pointed out in the diff:
- Flag manipulations instructions eg: `setf8` and the rest in this group and axflag are currently showing with throughput of 8 when it should be 1.
- SVE Duplicate, immediate and indexed form are showing throughput of 8 when it should be 2. I did not find tests in the patch, might be good to add some tests as well, but that's a NIT. The throughput needs fixing for these instructions, eg immediate: `dup     z3.d, #-256`; `dup     z2.h, #10`; or indexed: `dup     z4.s, z5.s[3]`

https://github.com/llvm/llvm-project/pull/185398