Feature request: Weigh binning confidence into GraphBin2 refinement

[schorlton]

Thanks for the great software! I'm wondering if it makes sense to weigh binning results with a confidence metric?

A toy example I'm thinking of is:
3 contigs total, all connected to each other on the assembly graph.
1 contig is 1 million bp and has bin ID 1
2 contigs are 100bp each and both have bin ID 2
My bins actually have meaning (eg. taxon identities or linkage to another dataset)

If I'm confident in the bin ID of contig 1, I don't actually want contigs 2 and 3 to override its bin ID. In fact, I may want the opposite!

Looking for feedback if this makes sense within the GraphBin2 framework, and if it would be possible to implement. Thanks again!