feat: add read_tabix and read_bam_references convenience functions

[pkerpedjiev]

• read_tabix: queries a BGZF tabix-indexed file for records in a genomic region, returning Arrow IPC bytes with chrom/start/end/raw columns. Accepts file paths and file-like objects; index can be a .tbi/.csi file path or file-like.

• read_bam_references: reads reference sequence names and lengths from a BAM file header, returning Arrow IPC bytes with name/length columns. Useful for building chromsizes without scanning the full file.

These functions maintain backward compatibility with clients that relied on the same API in earlier oxbow versions (e.g. HiGlass/clodius).

[nvictus]

Thanks, @pkerpedjiev !

For the bam references, doesn't this do the job already? The same property exists for all data sources.

chromsizes = ox.from_bam(...).chrom_sizes

For tabix/CSI-indexed files, the BED datasource already does what you need (from_bed(..., bed_schema="bed3+", index="...")), as long as chrom, start, end are the first 3 columns as required by the BED standard. With the last PR you can also apply custom type parsing to extended BED columns to do parsing on the Rust side (bed_schema=("bed3", {"foo": "int", "bar": "string"}).

DataSources are the preferred API for oxbow, returning iterators that expose record batches from Rust to Python with zero copy.

If I were to add something, I do think that we need a more generic BED-like TSV reader for the tabix use case where chrom, start, end are not the first 3 fields, as tabix allows this.

[pkerpedjiev]

Yeah, I think I can use both of those. I'll try to update the clodius PR with those changes and reopen and modify this if it doesn't work.