212 bug centroid nans sources not removed

[jeipollack]

Summary

Fixes an issue where sources with invalid centroid estimates (NaNs/Infs) were not removed, leading to downstream error during misalignment calculations using Centroid positions.

Introduces a robust batch filtering utility to enforce alignment and prevent silent data corruption.

Closes #212

What’s changed

• Added safe_batch utilities to compute validity masks from anchor arrays (e.g. centroids) and filter aligned datasets.
• Implemented safe_batch_builder to consistently apply masking across all sample-aligned arrays (images, SEDs, masks, object IDs, etc.).
• Enforced strict validation:
  • Raises on misaligned array lengths
  • Raises on invalid array type
  • Raises when all samples are invalid (prevents silent downstream failures)
• Extended support to sequence-based metadata (e.g. lists of object_id) to ensure proper alignment.
• Added logging utility to trace filtered samples.
• Added comprehensive unit tests covering:
  • Valid / partially invalid / fully invalid inputs
  • Alignment guarantees
  • Failure modes

How to test / verify

• Run unit tests:

pytest

• Specific checks:
  • Verify that datasets containing NaN centroids correctly remove affected sources.
  • Confirm that all output arrays remain aligned after filtering.
  • Confirm that fully invalid batches raise a ValueError.
  • Confirm that misaligned inputs raise errors.
• (Optional) Re-run a pipeline stage using real data where centroid failures previously occurred and verify:
  • No silent misalignment
  • Correct number of sources after filtering

Scope

Indicate the type of PR:

• Feature
• Bug fix
• Hotfix
• Documentation / process change
• Internal / refactor
• Release

Optionally, note if this PR is part of a larger milestone or set of related PRs.

Changelog

• Changelog fragment added (if applicable)

Reviewer Checklist

Reviewers should confirm the following before approving and merging:

• The PR targets the correct base branch (develop, or main for release PRs)
• The PR is assigned to the developer
• Appropriate labels are applied
• The PR is included in relevant projects and/or milestones
• Description clearly explains what has changed
• Issue references included, if applicable
• Code and documentation adhere to current standards (ruff)
• Documentation updates included, if relevant
• CI tests are passing
• All reviewer comments have been addressed

Next Steps / Notes (if applicable)