Making Unfused Loops Fused

[tanner-andrulis]

May want to track unfused loops going into the fusion step, because they may be connected to other fused loops when looking at Einsums with multiple data dependencies. For example, the following three pmappings may be fused:

           [B]        [B]
for m      for m      for m
[A]                   [A]

Right now the middle "for m" loop wouldn't be tracked because it isn't part of the compatibility of the middle Einsum.

Would it make sense to fuse these even if there's no "for m" loop in the middle Einsum?