Stop retrying kind 10011 on relays that permanently reject it

[saiy2k]

Context

Daily relay-directory-backfill in gr-prod repeatedly marks some relay/kind cursors as retry_later for permanent kind rejections, which adds churn every 06:00 UTC run without ever progressing.

Observed (stable across recent daily runs):

• closed:ERROR: bad req: filter validation failed: kind not allowed: 10011 on wss://relay.us.whitenoise.chat and wss://relay.eu.whitenoise.chat
• closed:kinds not supported on wss://relay.fountain.fm (kind 10011)

Transient failures (timeout, auth, etc.) should keep using retry_later.

Proposal

When a relay closes a REQ because the kind is permanently unsupported (kind not allowed / kinds not supported):

1. Checkpoint the cursor as terminal (status: "unsupported", completed: true) instead of retry_later
2. Skip that cursor on resume (same terminal path as complete)
3. Preserve the original lastReason for observability

Acceptance

• Permanent kind rejects no longer produce daily retry_later for the relays above
• Timeouts / other closed reasons still retry
• Next daily run after deploy shows unsupported (or skip) rather than re-querying those cursors