v0.3.1

Added

• Two new ReaPack repository configuration steps from the REAPER
  Accessible team: REAPER Accessible (EN) (reaperaccessible/rap_en,
  English-language resources, recommended) and REAPER Accessible (FR)
  (reaperaccessible/rap_fr, French-language resources, optional — offered
  but not pre-selected, since its content is French-specific). Like the
  existing REAPER Accessibility step, they're offered after install when
  ReaPack is present and wire the repositories into reapack.ini.

Changed

• Installations are faster: package downloads now run in parallel (up to
  three at a time), and each package installs as soon as its own download
  finishes instead of waiting for every download to complete. In a typical
  setup, REAPER, OSARA, and SWS are already installed while FFmpeg's large
  archive is still downloading. Installs themselves still run one at a
  time and in the usual order, so nothing changes about elevation prompts
  or the order packages land on disk. The progress page now tracks several
  simultaneous downloads: the bar combines their progress, and the status
  line summarizes them ("Downloading 3 packages…") while an install is not
  running. If something fails partway through, packages that already
  installed keep their receipts, so RABBIT still knows about them
  afterwards.

Fixed

• After the online version check, the packages page now places screen-reader
  focus on the top-most entry of the package list (the "Packages" group
  header) instead of landing on an arbitrary row a few items down. Going
  Back to the page still keeps your previous position.
• Large downloads no longer fail on slow or briefly stalling servers.
  Package downloads (and RABBIT's own self-update download) previously
  inherited a 30-second network timeout, so a download whose connection
  stalled for longer — routine for FFmpeg's ~390 MB archive from the busy
  gyan.dev server — was aborted with a cryptic "I/O error … error decoding
  response body" and took the whole installation with it. Downloads now
  tolerate stalls of up to a minute, and an interrupted connection is
  retried up to three times, resuming from the bytes already received when
  the server supports it (verified so a resumed file can never mix two
  different upstream versions) instead of starting over. If the connection
  keeps dropping, the error now clearly names the download URL and suggests
  checking the internet connection, rather than pointing at a temp file on
  disk.