Fix --resume manifest truncation bug

[ranman]

Summary

• Root cause: The resume code path in runner.py never updated the manifest metadata header — ManifestMeta was only constructed in the fresh-scan branch. On resume, the header was stale and records could be lost if create_manifest() was reached instead of write_meta().
• Hoisted ManifestMeta construction above the if/else and added a write_meta() call in the resume branch to atomically update the header while preserving all existing image records.
• Expanded pipeline tests from 4 to 8: no-duplicate appends, record preservation, partial-run resume, force truncation, nonexistent manifest, and metadata refresh.

Test plan

• uv run pytest — all 69 tests pass
• uv run ruff check src/ tests/ — clean
• uv run mypy src/imgeda/ — clean

🤖 Generated with Claude Code

[amazon-q-developer]

Summary

This PR successfully fixes the manifest truncation bug by hoisting ManifestMeta construction and adding write_meta() in the resume path. The test coverage expansion from 4 to 8 tests is excellent and validates the fix.

Key Changes:

• ✅ Hoisted ManifestMeta construction above resume/fresh-scan branches
• ✅ Added write_meta() call in resume path to update header without truncation
• ✅ Comprehensive test coverage for resume scenarios

Review Items:

• 1 comment regarding metadata timestamp behavior during resume operations

All tests pass and the core bug fix is correct. Please review the comment regarding created_at timestamp handling.

---

You can now have the agent implement changes and create commits directly on your pull request's source branch. Simply comment with /q followed by your request in natural language to ask the agent to make changes.

[amazon-q-developer]

🛑 Logic Error: Setting created_at to current time on resume overwrites the original scan creation timestamp. During resume, the metadata should preserve the original created_at from the existing manifest to maintain accurate scan history. Currently, every resume resets this timestamp to "now", losing the original creation time.

Suggested change

	meta = ManifestMeta(
	input_dir=os.path.abspath(input_dir),
	total_files=total_discovered,
	created_at=datetime.now(timezone.utc).isoformat(),
	settings={
	"workers": config.workers,
	"include_hashes": config.include_hashes,
	"skip_pixel_stats": config.skip_pixel_stats,
	"artifact_threshold": config.artifact_threshold,
	"dark_threshold": config.dark_threshold,
	"overexposed_threshold": config.overexposed_threshold,
	},
	)
	meta = ManifestMeta(
	input_dir=os.path.abspath(input_dir),
	total_files=total_discovered,
	created_at=datetime.now(timezone.utc).isoformat(),
	settings={
	"workers": config.workers,
	"include_hashes": config.include_hashes,
	"skip_pixel_stats": config.skip_pixel_stats,
	"artifact_threshold": config.artifact_threshold,
	"dark_threshold": config.dark_threshold,
	"overexposed_threshold": config.overexposed_threshold,
	},
	)
	if config.resume and not config.force and output.exists():
	processed_set, existing_records = load_processed_set(output_path)
	already_processed = len(existing_records)
	pending = filter_pending(all_images, processed_set)
	# Preserve original creation timestamp during resume
	from imgeda.io.manifest_io import read_manifest
	existing_meta, _ = read_manifest(output_path)
	if existing_meta and existing_meta.created_at:
	meta.created_at = existing_meta.created_at
	# Update metadata header without truncating existing records
	write_meta(output_path, meta)