Make file sorting optional in shard_from_audio_dir for large corpora

[Copilot]

Addresses performance feedback from #40: sorted(input_dir.rglob(...)) materializes and sorts all file paths up-front, causing unnecessary memory/time overhead for large audio datasets.

Changes

• Add sort_files parameter (default: False) to shard_from_audio_dir
  • When False: iterates lazily over files without sorting (memory efficient)
  • When True: materializes full file list and sorts (deterministic ordering)
• Update test coverage: add test_sort_files_parameter to verify both behaviors

Example

# Default: efficient for large corpora, non-deterministic order
shard_from_audio_dir("./audio", "./shards")

# Opt-in to deterministic ordering when needed
shard_from_audio_dir("./audio", "./shards", sort_files=True)

The generator-based approach eliminates the memory spike from loading all paths before processing begins.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.