feat(meta): seed 1013 tickets into pql changelog (pql migration phase 2)

One-way ticket data migration from the legacy settledreach.db into pql's
git-tracked changelog. Establishes the T-N == #N id bijection (old #440 -> T-440)
so thousands of #NNN git-commit references stay a trivial mapping.

Migrated (verified by full `pql plan rebuild` from changelog):
  - 1013 tickets  (status + type distributions match source exactly)
  - 481  dependencies
  - 16   history rows (deterministic content hash so ON CONFLICT(hash) dedups)
  - 90   labels = 48 source + 42 milestone-derived
           (active "Phase 4" milestone -> phase:4; "Process Rewire" ->
            milestone:process-rewire; milestone_deps was empty/vestigial)

Transforms: ids T-prefixed, sprint_id dropped (legacy/archival), decision_ref
and the one 'server,client' comma-team (#575) kept verbatim, deleted_at NULL,
canonical_version 1. Seeded rows carry hash=NULL on tickets/deps/labels — proven
safe: PK-based ON CONFLICT, updated_at drives LWW, and replay is idempotent.

Key finding that dissolves pql-requirements item #1 ("Critical"): seeding via
direct-INSERT does NOT require explicit ticket ids from `ticket new`. The id
counter derives from max(id), so after seeding T-1..T-1021 the next native
`pql ticket new` mints T-1022 — no recycling, no collision. The blocker only
applied to the `ticket new --id` path we never use for bulk seeding.

Scaffolding: .pql/changelog/<table>/0000-schema.sql (canonical pql 1.6.2 schema,
byte-identical across the 4 table dirs) + .gitattributes union-merge driver so
changelog SQL never produces binary-style merge conflicts. Seed is reproducible
via tooling/pql-migrate/seed_tickets.py (read-only on the source, idempotent).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.gitattributes

@@ -0,0 +1 @@
+.pql/changelog/*.sql merge=union