[rocm-libraries] ROCm/rocm-libraries#7725 (commit eef7e12)

[GFX1250][CK_TILE] Add scale16 warp gemm unit tests

## Summary
- Add scale16 WMMA intrinsic overloads and int64_t forwarding to warp
gemm layers for gfx1250
- Add comprehensive wave-level unit tests for scale16 warp gemm
(16x16x128 and 32x32x128 tile sizes)
- Test all fp8/bf8 type combinations and TransposeC variants
- Fix WarpGemm wrapper for non-uniform scale16 configurations

Stacked on #7724 (FillUniformScaleDistribution / MX GEMM scale init).
Pipeline enablement follows in the next PR.