fix(parse): preserve strict column mismatch reasons

[11suixing11]

Summary

• Preserve withHeaders.reason when strict column handling marks a parsed row invalid.
• Add an issue regression test for header transformation plus strict column handling with both too many and too few columns.
• Exercise the public data-invalid event and verify rows, reasons, emitted data, and final row count.

Fixes #772.

Why

HeaderTransformer already creates the Column header mismatch... reason, but CsvParserStream.transformRow() dropped that reason before emitting data-invalid. Consumers listening for the third data-invalid argument therefore received undefined even though the mismatch reason was known.

There was an earlier closed PR (#1112) for the same issue, but current main still reproduces the behavior. This version adds focused coverage for the exact reported path.

Verification

• corepack pnpm exec jest packages/parse/__tests__/issues/issue772.spec.ts --runInBand
• corepack pnpm test (38 suites, 604 tests, and all examples)
• corepack pnpm run build (7 workspace projects)
• corepack pnpm run format:check
• git diff --check

[11suixing11]

I re-ran this branch after the upstream rebase.

Local verification:

• pnpm exec jest packages/parse/__tests__/issues/issue772.spec.ts --runInBand — 1 test passed
• git diff --check — clean

GitHub currently reports the Test and dependency-review workflows as action_required with no jobs, and the PR is blocked on required review. No further code change is needed from my side unless the maintainers identify a different compatibility concern.

[11suixing11]

Additional independent verification for this PR:

• Clean frozen-lockfile install completed.
• Workspace build passed.
• After the build, lint passed.
• Jest passed: 38 suites / 604 tests.
• Workspace examples passed.

A direct pnpm test before the build still reaches the repository's existing workspace-resolution lint baseline; the build-first sequence above completes cleanly. No code changes were needed in this PR.