Skip comment fetches for status-filtered posts; drop HelpScout inter-page sleep

[dkships]

Two speedups for the slow fetch paths.

ProductLift: filter by status before fetching comments. get_feature_requests fetched comments for every post (250ms sleep + one API call each), then applied the status filter afterwards. The filter now runs inside fetchFeatureRequests before the comment loop, so filtered-out posts cost nothing. Same match semantics as before: status name compared case-insensitively, posts with no status excluded. On a portal with 200 posts and 20 matching, that's 180 fewer API calls and ~45s less sleeping per call.

HelpScout: remove the inter-page sleep in fetchConversations. The 250ms sleep was the effective pacer (~90-130 req/min with network time). Removing it lets pages run at network speed; the existing 170/min sliding-window cap and 429 retry/backoff keep us under HelpScout's documented 200/min limit. fetchMailboxes keeps its sleep (PAGE_DELAY_MS is still used there).

Behavior note: pii_categories_redacted may shrink when a status filter is set. It now reflects only PII found in the posts/comments actually returned, not in posts the filter would have discarded anyway. The returned data is scrubbed exactly as before.

Tests: two new vitest cases mock fetch and assert no /comments request fires for posts that fail the filter, and that unfiltered calls still fetch comments for every post. 66/66 passing.

🤖 Generated with Claude Code