Add "pin" attribute and re-solve motile with pins

[cmalinmayor]

Include:

• node selected (or not)
• edge selected (or not)
• mother cell (or not)
• endpoint (or not)

Mother and endpoint require motile implementations, but node and edge selected are already implemented in motile. We need to figure out the UI for it and add the constraint to the solver.

[cmalinmayor]

The user will be editing the "solution" labels. We need to incoporate those edits back into the input to motile.

Option number 1: feed the solution labels back into motile as input. Challenges: All the node ids will be different than in the original, because they will be re-computed from the relabeled labels.

Option number 2: Feed the solution Tracks back into motile, and add candidate edges back in (basically, skip the step of getting nodes from segmentation). Challenges: Need to re compute the IOU for all the new edges. This is the same as would happen in option 1, so it isn't worse. * This is the correct short term solution, I think. We look in the Tracks list for Tracks we can use as input. Tracks will also hold the pinned information. Deleted nodes are just gone, so no pinning. Deleted edges need to be kept track of, and selected nodes and edges.

Option 3: Display the candidate graph and have the user edit that directly. This is the "correct" long term solution. It incorporates with our idea of groups as subsets of a trackgraph, where the solution is a group (and if the layer data is separated from the Tracks.segmentation, the layer can relabel the segmentation based on the solution group ID). However, the track graph now isn't following our assumptions of binary tree, and this will be tough to visualize. Basically, we want a "view" into the candidate graph, so they stay synchonized.