[E09] Web page reference fetcher service

[IanMayo]

Epic

Part of E09: Provenance Graph

Problem

When an analyst captures a reference value from a web page (URL + XPath/CSS selector), Debrief needs to be able to re-fetch that value later and determine whether it has changed. There is no service to perform this fetch-and-compare operation.

Proposed Solution

Python service that:

1. Accepts a reference source record (URL + selector)
2. Fetches the web page
3. Applies the XPath/CSS selector to extract the current value
4. Compares against the reference value at capture
5. Returns the appropriate currency state:
   • current — fetched value matches captured value
   • changed — fetched value differs
   • unavailable — source could not be reached
   • source-structure-changed — source fetched but selector returned no match

This is the initial source type implementation; the service interface should be extensible for REST API, local file, and cross-session sources later.

Success Criteria

• Service correctly determines all 4 currency states for web page sources
• Handles network errors gracefully (returns unavailable)
• Handles DOM restructuring (returns source-structure-changed)
• Unit tests cover all currency state transitions
• Exposed via MCP following existing service patterns

Dependencies

Requires #145 (reference data source schema)

Complexity

Medium