Cody v0.56.0

Release 0.56.0

Long unattended runs keep going overnight.

• A chat whose run gets stuck is restarted by Cody. That covers a tool call
  that never answers, an engine that stops responding, or an hour with no
  progress. Cody then tells the agent what happened, so it checks what
  finished and carries on.
• If a chat's engine crashes mid-run, Cody restarts it after 30 seconds.
• If Cody's server or container restarts in the middle of a run, the run is
  picked up again when it comes back (if it was active in the last 6 hours).
• At most 3 restarts per chat in 12 hours, then Cody stops and tells you. Every
  restart sends a push notification if notifications are set up.
  CODY_RUN_RECOVERY=0 turns all of this off.
• Scheduled messages are no longer lost when they wait behind a long reply.
  The row says "waiting for the current reply to finish…", and the message is
  sent again if the engine or the server stops before reading it.
• A chat can no longer split into two copies with an engine on each when its
  engine is replaced.
• The server log now says why a running chat's engine was stopped.
• Unraid: the template raises the container's process limit to 16384 (Docker's
  default is 2048, which long runs exceed). Add --pids-limit 16384 to Extra
  Parameters on an existing install; the container log warns when it is low.

Full Changelog: v0.55.0...v0.56.0