v0.4.3

Fixed

• CSV/TSV parsing uses Polars consistently. FlashLFQ io.add_quant, metadata and SDRF
  now share the search-result readers' text parser, avoiding the pandas decimal-parsing
  differences observed between macOS and Linux before add_quant casts to float32. Matrix
  storage dtypes, provenance schema and the msmu-v1 hash policy are unchanged. Corrected
  numeric values can produce different output hashes when replaying an older import; use
  the recorded original environment or regenerate the reference with the new parser.
  SDRF strings, empty values and duplicate/unnamed headers are preserved. General metadata
  follows Polars inference, including its handling of whitespace-padded numbers and blank rows.

Merged pull requests

What's Changed

• BID-345 Unify CSV and TSV parsing with Polars by @jlthepi in #60
• BID-346 Release MSMU 0.4.3 by @jlthepi in #61
• BID-346 Release MSMU 0.4.3 by @jlthepi in #62

Full Changelog: v0.4.2...v0.4.3