Extend CK gemm/sdpa support to gfx950 (#45)

Update CK for gfx950 (#49)

(cherry picked from commit 8ccfc47)
(cherry picked from commit b5d5987)

Author: alugorey

aten/src/ATen/Context.cpp

@@ -419,7 +419,7 @@ void Context::setROCmFAPreferredBackend(at::ROCmFABackend b) {
   if(b == at::ROCmFABackend::Ck) {
     static const bool ck_unsupported = []() {
       static const std::vector<std::string> archs = {
-          "gfx90a",  "gfx942"
+          "gfx90a",  "gfx942", "gfx950"
       };
       for (auto index: c10::irange(detail::getCUDAHooks().deviceCount())) {
         if (!detail::getCUDAHooks().isGPUArch(archs, index)) {

aten/src/ATen/native/transformers/hip/flash_attn/ck/fmha_bwd.hpp

@@ -453,4 +453,5 @@ struct fmha_bwd_traits
     bool is_deterministic;
     // TODO: padding check is inside this api
 };
+template <int Version = 2>
 float fmha_bwd(fmha_bwd_traits, fmha_bwd_args, const ck_tile::stream_config&);