fix(constraints): consider numeric/datetime extreme value clipping

[michdr]

Note

Imposes training-bound clipping for Inequality constraints and wires workspace context through constraint translation and generation.

• Add workspace_dir plumbed through ConstraintTranslator and _create_constraint_handler to InequalityHandler
• Enhance InequalityHandler to clip high_column to training stats min/max via Workspace.tgt_stats for both numeric and datetime; add helpers _extract_min_max_from_stats and _clip_to_training_bounds
• Update data generation step to pass workspace_dir and post-process synthetic parquet with the updated translator
• Adjust e2e test: assert clipping for ELAPSED_TIME and ARRIVAL_TIME, relax strict mean-time check (commented) and require most time deltas within 2–3 hours

^{Written by Cursor Bugbot for commit 8b41828. This will update automatically on new commits. Configure here.}

[cursor]

This is the final PR Bugbot will review for you during this billing cycle

Your free Bugbot reviews will reset on January 17

Details

Your team is on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle for each member of your team.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

[cursor]

Clipping can be undone by constraint correction

Medium Severity

The _clip_to_training_bounds method clips high_column values to training bounds, then at lines 190-195 sets high = low if the clipping caused a constraint violation. When low_column values exceed the training max for high_column, this correction pushes high back outside the training bounds, effectively undoing the clipping. This can cause test assertions expecting values within bounds to fail in edge cases where the model generates extreme low_column values.