onnxruntime

e7c9a6c2 - [webgpu] make DP4AMatMulNBitsSmallMProgram shader template (#25025)

[webgpu] make DP4AMatMulNBitsSmallMProgram shader template (#25025) ### Description This commit refactors the `DP4AMatMulNBitsSmallMProgram` to allow both `tile_size_k_vec` and `tile_size` to be configured. This change allows more flexibility for performance tuning without altering the core shader functionality. There is no functional change in this commit. ### Motivation and Context This is a preparatory change to enable `DP4AMatMulNBitsSmallMProgram` performance optimization work in subsequent commits. --------- Co-authored-by: Jiajia Qin <jiajiaqin@microsoft.com>