Currently, we generate a container class LongReadBAMToBreakpointMetadata that stores information such as Chimeric Alignments + Copy-Number segments from pre-processing our aligned BAM file and .cns seeds from CNVkit segmentation. We iteratively mutate components of this class to contain different transformed versions of these inputs in confusing + brittle ways. We also rely on metadata provided by the container class when attempting to generate the ILP model corresponding to a single BreakpointGraph, which shouldn't be necessary. After we do the initial amplified interval search and refinement, we should be able to generate self-contained BreakpointGraphs that can be processed independently.

In terms of the final outputted _graph.txt file, this PR adds PathConstraint + AmplifiedInterval information for each BreakpointGraph. This allows us to use just a *_graph.txt file with the cycle entrypoint to generate/solve an ILP model (no additional usage of a BAM or CNV seed file).