fix(migrations): bump v0.12.0 backfill timeout from 10min to 30min

[orendi84]

Problem

phaseCBackfillLinks and phaseDBackfillTimeline in src/commands/migrations/v0_12_0.ts each wrap gbrain extract {links,timeline} --source db in an execSync call with timeout: 600_000 (10 minutes).

On a brain of non-trivial size that is too tight. On my 13,002-page brain, each extraction takes ~11 minutes - right past the cutoff. Both phases return failed, the orchestrator marks the migration partial, and the daily post-upgrade hook keeps retrying the same thing and failing the same way every day.

Fix

Bump both timeouts from 600_000 (10 min) to 1_800_000 (30 min).

At the observed rate of ~50 sec/1k pages this covers brains up to ~35k pages; larger brains will need a proportional extension, but 30 min is a far safer default than 10 min for any non-trivial corpus.

Verification

On a 13,002-page brain (127,093 chunks, all embedded):

Before the fix: migration consistently finished as partial, both backfill phases timed out at ~10 min with no output written.

After the fix: migration completes cleanly in ~22 min total. Both backfill phases run to completion (~11 min each), produce their summary lines, and verify phase runs after. Migration status: complete.

Scope

One-line change to two lines in one file. No new imports, no test changes needed (the existing test/migrate.test.ts doesn't exercise real execSync against prod-sized data so the timeout path isn't covered by tests today - consider that a follow-up).