feat: Support max_files filtering on cached metadata

[MoAly98]

Problem

When metadata is generated and cached, the max_files config option is only applied during generation, not when loading from cache. This means if you generate metadata with all files but later want to run a quick test with fewer files, you're stuck with the full dataset.

Current behavior:

• processes filter: Works on cached metadata
• max_files filter: Only works during generation

Example

# First run: generate metadata with all files
config["datasets"]["max_files"] = None
config["general"]["run_metadata_generation"] = True
# ... generates metadata/fileset.json with 1000 files per dataset

# Second run: want to test with just 10 files
config["datasets"]["max_files"] = 10
config["general"]["run_metadata_generation"] = False  # use cached
# ... but workitems still contain all 1000 files!

Expected behavior

When loading cached metadata, max_files should filter the workitems to only include the first N files per dataset, matching what would happen if metadata was regenerated with that setting.

Suggested approach

Add a filter_by_max_files() utility (similar to existing filter_by_process() in utils/filters.py) and apply it either:

• In DatasetMetadataManager._load_existing_metadata()
• Or in run_processor_workflow() alongside the existing processes filtering

This would allow quick iteration on cached metadata without regenerating.