fix(genesis-writer): emit release_date as RFC3339, not Postgres text

[rickyrombo]

Unreleased tracks are published early

The tracks query selected release_date::text, which yields Postgres's own
2026-09-06 22:06:00. The indexer's parseReleaseDate accepts RFC3339,
RFC3339Nano and Mon Jan 02 2006 15:04:05 GMT-0700 — none of which match — so
releaseDateOrDefault silently fell back to block time (the source row's created_at).

That is not a cosmetic date difference. One format mismatch produces three symptoms:

1. release_date is rewritten from the future into the past
2. ScheduledReleasePublisher then matches is_unlisted = true AND is_scheduled_release = true AND release_date <= now() and sets is_unlisted = false
3. …and updated_at = now(), which is why some rows carried the replay wall-clock

Measured on the 2026-08-07 snapshot

Unlisted tracks with a future release date, rewritten	372
…of those, published early (is_unlisted flipped)	368
Tracks still holding a future release date	498 → 136

Verified end to end — track 2073330890

source        release_date 2026-09-06 22:06:00   is_unlisted true
migration tx  release_date 2026-09-06 22:06:00   is_unlisted true    <- writer carried both
indexed row   release_date 2025-08-10 20:57:26   is_unlisted false   <- exactly its created_at

Fix

Scan release_date as a timestamp and format it RFC3339. Every other timestamp
the writer emits already goes out as RFC3339 — this one was the outlier.

TestReleaseDateIsEmittedInAnAcceptedLayout asserts the output parses under the
layout the indexer actually uses, and that the Postgres text shape cannot come back.

🤖 Generated with Claude Code