Batch directory ingestion

[ai-mindset]

Add batch directory ingestion for meeting archives

Government teams generate dozens of recordings per month. Processing one by
one is not viable.

• Add mix wiki.add_dir /path/to/recordings/ that recursively discovers
  supported media files
• Process each through the existing queue/chunker pipeline
• Report progress per file (success/failure with reason)

This enables building a knowledge base from an entire team's meeting archive
in a single command.