parquet_encode: format timestamp columns as RFC3339Nano strings for compatiblity

[josephwoodward]

No description provided.

[claude]

Commits
LGTM

Review
One issue: the existing unit test for TYPE_DATETIME in schema_test.go was not updated to match the new RFC3339Nano string output, so task test:unit will fail.

1. schema_test.go:132-137 — the "time.Time passthrough for datetime" case still expects time.Date(...), but mapMessageColumn now returns "2024-12-10T15:30:45Z". Test name and expected value need updating; consider adding coverage for sub-second precision and non-UTC zones.

[claude]

Commits
LGTM

Review
Focused change converting MySQL CDC time.Time datetime/timestamp values to RFC3339Nano strings for downstream parquet compatibility, with the existing string-passthrough nil behavior preserved for zero/invalid timestamps. Tests cover sub-second precision and non-UTC zones.

LGTM

[claude]

Commits
LGTM

Review
Clean refactor of encodingCoercionVisitor.visitLeaf that adds a time.Time case alongside the existing string case for TIMESTAMP encoding. The ts and unit-switch are correctly hoisted so both input types share the same conversion path, and the error message is updated to reflect the new accepted type.

LGTM