Railbase core v0.5.14

Use all CPU cores on full-collection scans

Full-collection reads — the REST record-list handler and the storage.Find/Count helpers — now decode CBOR document bodies in parallel across all available cores, while applying predicates sequentially so results stay in key order. On an 8-core machine a 20k-document scan is ~3.4× faster; small collections and single-core hosts are unchanged (no regression).

• New optional RAILBASE_MAXPROCS / --maxprocs to pin the OS-thread cap (default 0 = auto). Go 1.25+ already detects the CPU count and honours a container's cgroup quota, so auto is correct almost everywhere — this only lets you deliberately pin it.
• Startup now logs detected logical cores + effective GOMAXPROCS.

No config or on-disk format changes — drop-in upgrade from v0.5.13.

Verify a download: shasum -a 256 -c checksums.txt.