Example contraction splitk (#430)

* start split k

* add base device class

* add example after merge develop

* add gridwise gemm

* add b matrix split k

* split=1

* change name for kb

* not bias result right

* bias only add once

* fix register spill

* regular code

* add fp32 example

* fix for 64bit index

* fix CheckValidity of gridwise

include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp

@@ -209,6 +209,8 @@ struct BlockToCTileMap_KSplit_M00_N0_M01Adapt
         const auto M0 = math::integer_divide_ceil(c_grid_desc_m_n_.GetLength(I0), MPerBlock);
         const auto N0 = math::integer_divide_ceil(c_grid_desc_m_n_.GetLength(I1), NPerBlock);
 
+        block_1d_id = block_1d_id % (M0 * N0 * KSplit_); // hide groups
+
         const index_t idx_ksplit = block_1d_id / (M0 * N0);
         block_1d_id              = block_1d_id % (M0 * N0);