refactor: use Parquet metadata for row counts in validate

[shaypal5]

Summary

• Replace pd.read_parquet() with pq.read_metadata().num_rows in _check_task_splits() for row count validation — avoids loading full Parquet data
• Replace pd.read_parquet(path, columns=[]) with pq.read_schema().names in _check_leakage() for column name checks — reads only the schema footer
• Add 3 new tests: metadata/data consistency, task split row count mismatch, leakage column detection via schema

Note: _check_tables() still loads full DataFrames because they are needed downstream for FK integrity checks.

Closes #17

Test plan

• All 757 tests pass
• Ruff lint clean
• New test test_task_split_row_count_mismatch verifies error detection via metadata
• New test test_leakage_detects_extra_columns verifies column detection via pq.read_schema()
• New test test_task_split_metadata_matches_data verifies metadata row counts match actual data

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Refactors bundle validation to use Parquet footer metadata/schema for lightweight checks, reducing memory usage and speeding up validate on larger bundles.

Changes:

• Use pyarrow.parquet.read_metadata(...).num_rows for task split row-count validation instead of loading full DataFrames.
• Use pyarrow.parquet.read_schema(...).names for leakage column detection instead of pd.read_parquet(..., columns=[]).
• Add tests covering metadata/data consistency, task split row-count mismatch detection, and leakage detection via schema.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
leadforge/validation/bundle_checks.py	Switches task split row-count checks to Parquet metadata and leakage checks to Parquet schema reads.
tests/validation/test_bundle_checks.py	Adds regression tests for metadata row counts, task split row mismatch detection, and schema-based leakage detection.
.agent-plan.md	Documents completion of the Parquet-metadata validation refactor and associated tests.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

pr-agent-context report:

No unresolved review comments, failing checks, or actionable patch coverage gaps were found on PR #39 in repository https://github.com/leadforge-dev/leadforge. Treat this PR as all clear unless new signals appear.

Run metadata:

Tool ref: v4
Tool version: 4.0.21
Trigger: commit pushed
Workflow run: 25223507963 attempt 1
Comment timestamp: 2026-05-01T16:51:11.951535+00:00
PR head commit: 052ee9e010d3e6310288a6a184323b13bb5e6f42