Refactor clonotype_imbalance

[grst]

Analogous to #156, at some point

• [ ] make volcano and dotplot a different function
• [ ] make clear what is computed by the tool and what on-the-fly
• [ ] more compact representation for dotplot, potentially a heatmap could work.
• [ ] speed up:
  • use implementation of fisher's test that approximates using chi-square when sample size is large enough
  • should be trivial to parallelize

[grst]

Potentially, clonotype_imbalance will be made obsolte by clonotype_modularity. For now, I marked clonotype_imbalance as deprecated with a note asking the user to switch to clonotype modularity, or drop an issue if that doesn't fit their use-case.

Let's see how that goes, in the meanwhile, clonotype_imbalance will remain as is.