feat: add classify, crawls, export CLI commands

Unify the full pipeline under the corpus CLI:
- `corpus crawls` lists available CDX-filtered crawls from R2
- `corpus classify` wraps Python ML classification as subprocess
- `corpus export` wraps HuggingFace parquet export via uv
- `corpus status` now shows LLM classification breakdown by type

Rewrite README to reflect the complete pipeline flow.

Author: caiopizzol