Add optional BINSEQ/CBQ input support for faster sequencing I/O

[nick-youngblut]

At the Arc Institute, we use binseq for increased performance versus fastq files. It would be great to see binseq supported in STAR-suite as an optional input backend. Some features of binseq:

• Binary two-bit sequence encoding
• Support for single-end and paired-end reads
• Optional storage of qualities and headers
• A blocked/columnar layout
• Per-column ZSTD compression
• Potential lazy/partial decoding of fields that a workflow actually needs

CBQ input could provide up to 2.0x faster input ingestion vs native gzip FASTQ.

[lhhunghimself]

Looks easy enough to do - it's about time that we moved on to something that is faster to ingest. That is a real bottleneck.

[lhhunghimself]

CBQ input has now been implemented for STAR-suite and for Chromap-suite. For STAR-suite, the internal binary representations are essentially the same as those in CBQ and the I/O bound FLEX and feature processing modules really benefit from the parallel access. I hope the file format gains wide acceptance.