Fix docling example metadata extraction writing invalid JSON

[Croway]

Summary

The document-metadata-extractor route used docling:EXTRACT_STRUCTURED_DATA as the endpoint, but camel-docling only dispatches on the operation query parameter (or the CamelDoclingOperation header) - the URI path segment is descriptive only. Without an explicit operation=... parameter, the endpoint silently fell back to its CONVERT_TO_MARKDOWN default, so the "metadata" route was actually just converting to markdown and writing it into a .json file (invalid JSON).

• Both routes now pass an explicit operation=CONVERT_TO_MARKDOWN / operation=EXTRACT_STRUCTURED_DATA query parameter. The URI path segment itself was renamed to a lowercase, hyphenated label (docling:convert-to-markdown, docling:extract-structured-data) instead of repeating the operation name, to make clear it's just an endpoint id and the operation parameter is what actually selects the behavior.
• EXTRACT_STRUCTURED_DATA returns a DoclingDocument object, not JSON text (the outputFormat parameter isn't read by that operation, and there's no type converter for the object), so the route now marshals it to JSON via camel-jackson before writing it to file.

I've also filed the root cause against camel-docling itself: apache/camel#26102 (the docling:<operationId> URI path segment not being honored). This PR fixes the example regardless of when/whether that lands, since the explicit operation=... parameter and the JSON marshalling step are both needed either way.

Test plan

• Reproduced against a live docling-serve container (docker run -p 5001:5001 ghcr.io/docling-project/docling-serve:latest)
• Before: output/metadata/*.json contained markdown text, invalid JSON
• After: output/metadata/*.json contains valid, well-formed JSON (DoclingDocument schema)
• Verified the primary document-to-markdown-converter route still produces correct markdown output
• mvn clean package succeeds

🤖 Generated with Claude Code

https://claude.ai/code/session_01Hs6c6LpuL8ZZ3gaSXwt7CS