[clang-tools-extra] [compiler-rt] [llvm] [AMDGPU] Enable WMMA256bInsts + Wave32 for gfx1200/gfx1201 + SISchedule + TargetParser + Virtual FP4/MXFP4 (PR #201810)

[via cfe-commits]

clearnature wrote:

## Dependency chain for RDNA4 enablement

These changes are the foundation for library-level RDNA4 support:

**This PR (LLVM backend)** → provides the compiler intrinsics needed by:
- **ROCm/rocm-libraries#8089** — rocWMMA SWMMAC headers wrapping `__builtin_amdgcn_swmmac_*`

Without the WMMA256bInsts + Wave32 feature flags and SISchedule fixes here, the rocWMMA library headers cannot compile for gfx1200. The changes form a coordinated, two-layer RDNA4 enablement:

```
LLVM backend (this PR)     →  rocWMMA headers (PR #8089)  →  rocBLAS kernels (subsequent PR)
```

Both PRs are submitted and awaiting review.

https://github.com/llvm/llvm-project/pull/201810