Fix bug in map_fusion transformation
Four-lines bugfix and associated test case for map_fusion transformation.
Without this change, the test would fail in SDFG validation with error:
dace.sdfg.validation.InvalidSDFGEdgeError: Memlet data does not match source or destination data nodes) (at state state, edge __s0_n1None_n3IN_T[0] (V:None -> numeric:_inp))
@alexnick83 I can give you some context here. The problem addressed in this PR is different from the problem addressed in #1535. The problem addressed here blocks a delivery on gt4py main branch (for reference https://github.com/GridTools/gt4py/pull/1502), which simplifies the icon4py SDFGs.
Thank you @alexnick83 for reviewing this PR. I have limited the application of this fix to the known valid case, as you suggested.
However, I skipped this line:
if current_data is not None and current_data not in test_data and list(edge.data.subset) == [0]:
because it is always true: the current_data is set at line 463-464.