mmq.cu: tune mmq/rocblas switching for RDNA (llama/18537)

mmq.cu: tune mmq/rocblas switching for RDNA (llama/18537)

* Patch perf regression for mmq kernels in ROCm

recover performance regression for https://github.com/ggml-org/llama.cpp/issues/17917

* add n_experts branch like the cdna path

* mmq.cu: tune mmq/wmma switching for RDNA

* mmq.cu: move amd wmma mmq/wmma switching behind IS_RDNA3

* Update ggml/src/ggml-cuda/mmq.cu

Co-authored-by: Johannes Gäßler <redacted>
---------

Co-authored-by: Jiacheng (Jason) Chen <redacted>
Co-authored-by: jiachengjason <redacted>
Co-authored-by: Johannes Gäßler <redacted>