GitLab

Softmax fp32
propagate_constant fp64
layernorm fp32

Adjust Github Action flows to reduce the workload per commit by caching the docker image once so all jobs can use it

This resolves an edge case found in multiple huggingface models  in some cases the find_split_reshape matcher will match with reshape2, but vec_rsp will consist of reshape1 dims causing a shape mismatch error. Solution is to include rsp when checking all reshapes are the same.