Improved diplotype reconstruction

Briefly, this release adds new methods to reconstruct the most likely diplotype from the read alignment on the pangenome. The goal was to make this diplotype reconstruction more robust in cases where the reads are smaller or the depth lower.

One new approach implements an expectation-maximization-like algorithm to iteratively converge toward N module profiles. The previous approach, based on clustering the reads based on their similarity, has also been improved to focus on the most informative read overlaps. The diplotypes, generated for different values of N (and other parameters), are then evaluated and ranked, as before.

Other additions:

• Visualization of reads supporting "functional" CYP21A2 modules (here reads without any of the identified variants)
• New subcommands to get metrics about the reads parakit gafstats
• Bug fixes for visualization, annotation, diplotype reconstructions
• More documentation (although more is still needed)
• Call all variants (bubbles in the pangenome) instead of just known ClinVar variants
• Add deconstruct subcommand to list variants in the pangenome.
• Handle of GAF files produced by vg giraffe