[RFC] Mark tensor.dim and memref.dim as side effecting

As was noted by others, this was often discussed but never done. We concluded for a trait back then but an interface with some light-weight dynamic decisions is much better. Thanks for pushing this forward.

3 Likes