v1.0.1 — Smaller files, faster imports

What's new in v1.0.1:

• Files are now ~5x smaller than CSV — adaptive gzip compression on
each column block (falls back to raw when compression doesn't help)
• Faster imports — type guessing scans the full dataset (not just 3 rows)
and columns are converted in parallel
• Crash-safe writes — import writes to a temp file and renames, so a
crash never leaves a corrupted .tblx
• File integrity check — truncated or corrupted files fail immediately
with a clear error instead of producing garbage data

Under the hood: libtblx updated to v0.2.1 (reflection-free encoding,
precomputed column offsets for O(1) seeking).

No breaking changes to commands or CLI flags.

Full Changelog: v1.0.0...v1.0.1

v1.0.0

tblx — alpha release