v1.0.0

On file inputs, the single-threaded parser is now 1.1-4× faster than needletail, while the new multi-threaded parser is 3-6× faster than paraseq using 4 threads.

Breaking change: parsers now yield Result<Event, parser::Error> (instead of Event), enabling standard ? error propagation.

What's New

• Built-in multi-threaded parser (ParallelFastxParser) with multiple readers.
• Streaming multi-reader BGZF (.gz) and multi-frame Zstandard (.zst) parallel decompression with constant memory (~2 MB).
• Accelerated 2-bit bitpacking (PackedDNA and ColumnarDNA) for contiguous reads and multiline sequences.
• AVX-512 vector kernel for 512-bit scanning on supported x86 CPUs.
• Zero-overhead CRLF (\r\n) line ending support for Windows.
• reject_non_actg option, returning an ErrorKind::InvalidBase error at the first non-ACTG base.
• swap_dna_packed, swap_dna_columnar, swap_mask_non_actg, swap_mask_n and swap_packed_seq_vec to get the current sequence without allocating and reuse the given buffer.
• Calling a FastxParser getter disabled by the config (e.g. get_header with ignore_headers) is now a compile-time error.
• FastxParser is Send.
• Needletail and paraseq migration helpers.