Splitting up MCs should normalise port names

[ymherklotz]

Currently, running the dynamatic-to-graphiti.py script on gsum_many.dot, for example, will split up the MC into two MCs below:

"MC_a_a" [type="MC", in="in1:32*l0a", out="out1:0*e out2:32*l0d", bbID=0, memory="a", bbcount=0, stcount=0, ldcount=1];
"MC_a_b" [type="MC", in="in1:32*l1a", out="out1:32*l1d", bbID=0, memory="a", bbcount=0, stcount=0, ldcount=1];

However, MC_a_a has two outputs, one named out1 which connects to the end node, and one named out2. Currently we cannot support MCs that connect to the end node because all nodes need to be local to the body of the loop. Additionally, we assume that the MC port connecting to the load port is assumed to be out1.

We can either add more rewrites to the framework to handle MCs with different output names, however, it might be easier to handle this in the dynamatic-to-graphiti.py script. Additionally, we will have to add support to the script to remove connections to the end node for MCs.