feat: add support for trigger comments, trigger enabled/disabled state, and sequence comments

[ayusssmaan]

Four related gaps where pgschema silently ignored real schema properties,
causing silent drift between the declared model and the live database.

What changed

1. Trigger comments

pgschema had no awareness of COMMENT ON TRIGGER. A comment declared in
the model was never applied; an unwanted comment on the live DB was never
flagged as drift.

• ir/ir.go: add Comment string to TriggerDef
• ir/inspector.go + ir/queries/queries.sql.go: read pg_description
  for trigger OIDs
• internal/diff/trigger.go: diff and emit COMMENT ON TRIGGER ... ON ... IS '...'

2. Trigger enabled/disabled state

pgschema had no awareness of pg_trigger.tgenabled. A trigger disabled in
the model was never applied; a trigger enabled/disabled on the live DB was
never flagged as drift.

• ir/ir.go: add Enabled bool to TriggerDef
• ir/inspector.go + ir/queries/queries.sql.go: read tgenabled from
  pg_trigger
• internal/diff/trigger.go: diff and emit ALTER TABLE ENABLE/DISABLE TRIGGER

3. Sequence comments

pgschema had no awareness of COMMENT ON SEQUENCE. A sequence comment in
the model was never applied; an unwanted comment on the live DB was never
flagged as drift.

• ir/ir.go: add Comment string to SequenceDef
• ir/inspector.go + ir/queries/queries.sql.go: read pg_description
  for sequence OIDs
• internal/diff/sequence.go: diff and emit COMMENT ON SEQUENCE ... IS '...'

4. Fix: SERIAL sequence comment idempotency

When a table with BIGSERIAL/SERIAL columns was created for the first
time, pgschema correctly skipped emitting a standalone CREATE SEQUENCE
(because CREATE TABLE creates it implicitly). However, this also
accidentally skipped emitting any COMMENT ON SEQUENCE declared in the
model. On the second run the sequence existed in the live DB, the comment
diff fired, and all sequence comments were re-applied indefinitely on every
subsequent run.

• internal/diff/diff.go: track SERIAL-owned sequences that were
  skipped from addedSequences but carry a comment into a new
  addedSerialSeqComments list; emit their COMMENT ON SEQUENCE after all
  CREATE TABLE calls complete

Verified locally: two consecutive deploys against a freshly created wiser DB
— all 14 schemas report "No changes to apply" on the second run.

Tests

Four new fixtures, all passing:

• testdata/diff/comment/add_trigger_comment/
• testdata/diff/comment/add_sequence_comment/
• testdata/diff/comment/add_serial_sequence_comment_on_create/ ← idempotency fix
• testdata/diff/create_trigger/disable_trigger/

[greptile-apps]

Greptile Summary

This PR teaches pgschema to track more trigger and sequence metadata. The main changes are:

• Trigger comments are inspected, stored in IR, and diffed.
• Trigger enabled or disabled state is inspected, stored in IR, and diffed.
• Sequence comments are inspected, stored in IR, and diffed.
• SERIAL-owned sequence comments are emitted after implicit sequence creation.
• New fixtures cover trigger comments, sequence comments, serial sequence comments, and disabled triggers.

Confidence Score: 2/5

This should be fixed before merging.

• View trigger comment and enabled-state changes can be silently ignored.
• Disabled view triggers can be created as enabled.
• Added triggers on existing tables can miss both comments and disabled state.
• Existing serialized trigger models that omit enabled can be treated as disabled.

internal/diff/view.go, internal/diff/trigger.go, internal/diff/table.go, and ir/ir.go need follow-up.

Important Files Changed

Filename	Overview
internal/diff/view.go	View trigger diff and modify paths were not updated to preserve the new trigger metadata.
internal/diff/trigger.go	Table trigger creation handles comments and disabled state, but view trigger creation only handles comments.
internal/diff/table.go	Modified table triggers handle metadata changes, but newly added triggers on existing tables do not.
ir/ir.go	The new trigger enabled field has a zero-value default that conflicts with PostgreSQL's enabled-by-default behavior.

Comments Outside Diff (1)

1. internal/diff/table.go, line 1360-1371 (link)

   Added table trigger metadata skipped

   When a trigger is added to an existing table, this branch only collects the CREATE TRIGGER SQL. It does not emit the new COMMENT ON TRIGGER statement or the ALTER TABLE ... DISABLE TRIGGER statement for disabled triggers. A migration that adds a commented disabled trigger to an existing table will create an enabled trigger with no comment.

_{Reviews (1): Last reviewed commit: "fix: emit COMMENT ON SEQUENCE for SERIAL..." | Re-trigger Greptile}