add methods for MCP approach

[rjchallis]

Trying a different approach by using MCP. Have removed all UI for now in favour of to connect to users LLM chat client.

Summary by Sourcery

Replace the previous Flask-based UI and llama-index integration with a FastMCP-based GoaT server and minimal MCP-focused setup.

New Features:

• Introduce a FastMCP-powered GoaT MCP server exposing tools and resources for querying the GoaT API, including taxon existence checks and conditional count queries.

Enhancements:

• Add project metadata and dependencies via a new pyproject.toml for the GoaT MCP server.
• Document MCP-based setup and usage in the source README, including running the GoaT MCP server and example queries.
• Document a basic testing workflow for running intent-module tests against a local LLM and GoaT API.

Build:

• Simplify runtime dependencies to only fastmcp and httpx in requirements.txt, aligning with the new MCP-based architecture.

Documentation:

• Expand README with instructions for installing uv, creating a virtual environment, installing MCP dependencies, and connecting via an MCP client.
• Extend INSTALL guide with a testing section describing environment configuration and pytest usage.

Chores:

• Remove legacy agent, UI, query pipeline, and supporting scripts that are no longer needed with the MCP-based approach.

[sourcery-ai]

Sorry @rjchallis, your pull request is larger than the review limit of 150000 diff characters