Genvarloader

Feat

- write ZarrTracks in smaller chunks.
- write ZarrTracks in smaller chunks.

Fix

- remove wip vidx feature.
- relax numba version constraint
- rounding issues for setting fixed lengths on BED regions.
- more informative vcf record progress bar.

Feat

- improve record query performance by allowing nearest_nonoverlapping index adjustment to be computed on-the-fly in the weighted activity selection algorithm and thus also benefit from early stopping.
- more descriptive progress bar for constructing ZarrGenos from another file.
- add progress bar for reading VCF records.

Fix

- pylance update, catch possibly unbound variables.
- instead of failing, raise warning when encountering non-SNP, non-INDEL variants and skip them.

Fix

- more descriptive pbar when writing ZarrTracks from another reader.
- BigWigs, only keep contigs that are shared across all bigwigs.
- better error messages and catching cases for non-SNP, non-INDEL variants.
- avoid segfault caused when a TensorStore is forked to new processes.
- make ZarrTracks implement Reader protocol. feat: add NumpyGenos for in-memory representation. feat: better ZarrGenos.from_recs_genos progress bar.

Fix

- naming of .ends.gvl.arrow to .gvl.ends.arrow so file suffix parsing works correctly.

Fix

- remove pyd4 dependency, had unspectacular performance.