Gracefully handle input from multiple individuals in annotate

[dnil]

See e.g. nf-core/raredisease#429.

Several possibilities here:

1. actually use sample from RG to choose which blocks to use - effectively picking one individual
2. flag to select which sample to track
3. invent CSQ-like INFO tag for samples or use sample GT level output to store info, and actually make calls for each individual
4. make calls for each (affected) individual and annotate with smallest autozygous region subset only - useful for RD discovery, less so for e.g. corroborating deletions