[ROCm][inductor] Add TunableOp layout/scaled coverage suites (split diff) (#192148) (#192148)

Summary:

Adds the layout- and scaled-GEMM coverage suites -- `LayoutCoverageWildcardTest`, `BmmLayoutCoverageWildcardTest`, and `ScaledGemmTunableOpFP8Test` -- to `test_dynamic_tunable_ops.py`. These suites were split into a separate diff from the core dispatch-matrix suites so each diff stays under the 2000-LOC PR size limit; together the two diffs land the full test file. No test logic changed relative to the original single-commit version.

Test Plan: HIP_VISIBLE_DEVICES=0 buck2 test fbcode//mode/dev-nosan-amd-gpu --modifier rocm70 -c fbcode.enable_gpu_sections=true -c fbcode.triton_backend=amd -c fbcode.rocm_arch=mi350 fbcode//caffe2/test/inductor:dynamic_tunable_ops

Reviewed By: Ruishenl

Differential Revision: D114812043

Pull Request resolved: https://github.com/pytorch/pytorch/pull/192148
Approved by: https://github.com/mlazos