fix: address review comments on aggregate_sample_features refactor

[raylim]

Addresses all five review comments left on #115.

Changes

mussel/utils/feature_extract.py

• Input validation — validate each slide's arrays before processing: raises a ValueError with slide index and sample_id if feats.ndim != 2, coords shape is not (N, 2), or len(feats) != len(coords)
• Tighter type annotations — return type and internal variables changed from bare dict to dict[str, tuple[np.ndarray, np.ndarray]]

mussel/cli/aggregate_sample_features.py

• Fix memory regression — group slide indices by sample_id first, then read H5 files one sample at a time; peak memory now scales with the largest sample rather than the full input set

tests/mussel/cli/test_aggregate_sample_features.py

• test_aggregate_sample_features_multi_slide — assert concatenated content (not just shapes) via np.testing.assert_array_equal
• test_aggregate_sample_features_with_subsampling — assert reproducibility (same seed → identical outputs) and features/coords alignment after subsampling (using identifiable rows where feature value == coord value)
• test_aggregate_sample_features_invalid_shapes — new test covering the three new validation paths (1-D features, wrong coords shape, mismatched lengths)

[Copilot]

Pull request overview

Addresses prior review feedback on the aggregate_sample_features refactor by adding per-slide input validation, tightening typing, fixing a CLI memory regression, and strengthening tests around correctness and reproducibility.

Changes:

• Add per-slide validation in aggregate_sample_features with informative ValueErrors for invalid feature/coord shapes and mismatched lengths.
• Update the CLI to process inputs grouped by sample_id and load H5s one sample at a time to reduce peak memory usage.
• Expand tests to assert concatenated content (not just shapes), reproducibility with a fixed seed, feature/coord alignment after subsampling, and new invalid-shape coverage.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
mussel/utils/feature_extract.py	Adds per-slide input validation and tightens return type annotations for aggregated outputs.
mussel/cli/aggregate_sample_features.py	Refactors CLI workflow to group slides by sample and load/process/write one sample at a time to reduce peak memory.
tests/mussel/cli/test_aggregate_sample_features.py	Strengthens assertions for concatenation/subsampling and adds coverage for new validation error paths.

---

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

[Copilot]

OrderedDict isn’t needed here (project requires Python >=3.10, where plain dict preserves insertion order). Consider replacing groups: dict = collections.OrderedDict() with a typed plain dict (e.g., dict[str, list[int]] = {} or defaultdict(list)) and drop the collections import to reduce overhead and tighten typing (similar to mussel/utils/feature_extract.py:1989).