v0.21.0

• Default metapackage updated to GTDB R232 (S6.5.0-R232)
• pipe: Substantially improved logging, including progress bars for prefilter, hmmsearch and DIAMOND assignment steps. Thanks @thepatientwait.
• pipe: Parallelised DIAMOND taxonomic assignment for faster runs. Thanks @thepatientwait.
• pipe: Add optional --context-window if the entire query sequence is too large for output size.
• pipe: Add zstd input support in places, including with --read-chunk-size/--read-chunk-num and prokaryotic_fraction
• pipe: Restore v0.19.0 orfm+mux pipeline for genome input, restoring performance lost in v0.20
• pipe: Use DIAMOND --frameshift when assigning taxonomy
• pipe: --sra-files: Numerous fixes and refinements to chunked SRA extraction, including detection of kingfisher failures and stripping of .sra suffix from sample names
• pipe/renew: Bugfix for missing sequences with Lyrebird
• supplement: Add --output-matched-faa to output matched protein sequences, with marker name included in sequence IDs
• renew: Accept zipped gzip archive OTU tables as input
• lyrebird: Add extras/lyrebird_metapackage_creation snakemake workflow for building Lyrebird metapackages from scratch
• data: Report DOI when database download or acquisition fails
• Switch internal data handling from pandas to polars
• Updated citation for prokaryotic_fraction
• docs: Improved GlobDB documentation (#267)
• Add AGENTS.md and SKILL.md to assist agentic coding tools
• Assorted bug fixes, test improvements, and documentation updates

Thanks @thepatientwait, @rzhao-2, @AroneyS, @EisenRa and others for contributions and testing, and thanks to @MCeciC (#269) and @magicprotoss (#267) for raising issues.