[Epic] ai.data.gov; Natural Language Interface for Federal Open Data Discovery

### What
Establish Data.gov as the public-facing natural language search interface for federal open data, enabling users to find datasets using plain English queries rather than keyword search. This epic captures the foundational work required to be ready for integration with USAi when public-facing access becomes available.
### Why
Current catalog search requires users to know what they're looking for in advance. A natural language interface dramatically lowers the barrier for researchers, journalists, students, and the general public to discover and use federal data. Data.gov is uniquely positioned to deliver this given its mandate, its catalog API, and its relationship with the USAi team within TTS.
### Done When

- [ ] Metadata corpus is indexed and optimized for semantic retrieval
- [ ] Retrieval layer logic is built and tested against representative user queries
- [ ] Public-facing UI is designed and hosted on [cloud.gov](http://cloud.gov/)
- [ ] USAi API integration points are defined and stubbed, ready to activate when public-facing access is available
- [ ] Internal prototype demonstrates end-to-end query-to-dataset response

### Notes
USAi would be the intended LLM backend; public-facing access is not yet available but is theoretically on their roadmap
Natural partner for co-development story with USAi team -- Data.gov would be their first public-facing use case