v0.3.0

Added

• usdata pull fetches every source in a manifest and writes a lockfile pinning each asset's
  checksum and provenance; a later pull restores from the lockfile without re-resolving, and
  refuses (until --force) if the manifest changed. usdata verify reports missing or altered
  cached files. Python API: usdata.pull, usdata.verify.

• Datasets carry a status (available, stub, planned). search and info show it; fetching a
  planned dataset exits 3 with a clear message.

• Planned registry entries for USGS, NASA, EPA, FEMA, Census, USDA, and GOES so the roadmap is
  visible from usdata search and the docs.

• Provider metadata (name, homepage) in the registry; README gains a generated per-provider table.

• Datasets carry a domain (shared taxonomy across providers), since (version shipped) or
  target (planned phase or later). search and info show them; the roadmap lists
  datasets by target version, generated from the registry.

• 26 NOAA registry entries spanning 18 domains, each verified for anonymous access
  (except GHCN-Hourly, whose entry says its bulk path is unconfirmed), with target phases.
  The NOAA provider page gains a data-landscape table mapping domains to entries.

Changed

• usdata search hides planned datasets unless --planned is given.

Changed

• Dataset reference moved to docs/providers/: one page per provider with hand-written access notes and a generated dataset table, plus a generated index.