[chore]: simplify staging comment lifecycle and improve deploy status reliability

[ArtemLazarchuk]

Summary

• Comment lifecycle simplified — TryPostOrUpdateStagingCommentAsync now always
  deletes all prior marker comments and posts a fresh one. Removed forceNewComment
  parameter (was true in every non-trivial call anyway). Added retry loop in
  DeleteAllMarkerCommentsAsync so stale comments are reliably cleaned up.

• Immediate "queued" notifications — After enqueuing a deploy (new, redeploy,
  re-sync, webhook trigger) the PR comment is updated right away via
  TryNotifyDeployQueuedAsync, so contributors see progress instead of a stale
  "Deleted" or blank comment while the background worker starts.

• Deploy status resolution — Events and listing APIs are now polled in parallel
  and merged via MergeEventAndListingStatus (failure wins, then deployed, else
  pending). Removed the 5-minute "listing fallback" timer; listing is checked every
  tick, eliminating the lag window.

• Orphan service fix — DeployBranchAsync now tears down existing PR services
  before creating new ones, preventing leftover Sliplane services when a
  synchronize event triggers a fallback deploy alongside existing services.

• 404 on comment delete treated as success — DeleteIssueCommentAsync no longer
  returns false when the comment is already gone, removing spurious warnings.