Propagate existing attempt IDs into Q&A retrieval path (issue/short_id fallback, no attempt creation)

[Copilot]

Q&A runs were bypassing retrieval usage persistence because attempt_id could be None on the question path; as a result, retrieval_usage rows were skipped and log_retrieval_recorded() never emitted. This change ensures Q&A uses an existing attempt ID when available (issue-id first, short-id fallback) without synthesizing new attempts.

• Attempt ID resolution centralized before Q&A routing

  • Compute effective_attempt_id once per processor run:
    1. input.attempt_id
    2. tracker.get_attempt(source, issue_id)
    3. tracker.get_attempt_by_short_id(source, short_id)
  • Pass effective_attempt_id to answer_question_issue(...) and reuse consistently in retrieval-recording paths.

• Tracker API extension for short-id recovery

  • Added read-only lookup on FixAttemptTracker for (source, short_id) to recover pre-existing attempts when issue-id lookup misses.
  • Implemented backend query support so Discord-style IDs (e.g. DISCORD-15265329) resolve to the original attempt.

• No synthetic attempts

  • Preserved behavior when no existing attempt truly exists: no auto-creation in processor fallback logic.

• Targeted observability

  • Added debug logs for:
    • missing inbound attempt_id
    • recovered attempt_id source (issue_id vs short_id)
    • unresolved attempt after all fallbacks
  • This makes missing/recovered attempt propagation explicit during Q&A debugging.

let effective_attempt_id = input.attempt_id
    .or_else(|| self.tracker.get_attempt(&input.source_name, &input.issue.id).ok().flatten().map(|a| a.id))
    .or_else(|| self.tracker.get_attempt_by_short_id(&input.source_name, &input.issue.short_id).ok().flatten().map(|a| a.id));

if matches!(input.intent, Some(Intent::Question)) {
    return self.answer_question_issue(
        &input.issue,
        &input.resolution,
        &input.source_name,
        effective_attempt_id,
    ).await;
}