fix(api): report effective MarkItDown image OCR capability

[Sanderhoff-alt]

Summary

• Expose features.markitdown_image_ocr from GET /version so clients can distinguish general file upload support from usable MarkItDown image OCR.
• Report the capability only when file uploads and OCR are enabled, MarkItDown registered successfully, and the parser allowlist permits it.
• Regenerate the OpenAPI specification and Python, TypeScript, and Go client models, and update control-plane feature types.

Why

The optional MarkItDown OCR integration was controlled by dedicated server configuration, but /version only exposed file_upload_api. Clients could determine that the upload endpoint existed but could not determine whether image OCR was actually available. The missing capability flag made integrations rely on trial requests and conversion failures.

Impact

Clients can now check version.features.markitdown_image_ocr before offering or invoking MarkItDown image OCR. Existing deployments keep the default value false unless the complete runtime capability is available.

Validation

• API integration tests cover enabled, upload-disabled, allowlist-excluded, parser-unavailable, and OCR-disabled states.
• Python SDK version tests pass.
• TypeScript SDK version tests pass.
• Rust and Go clients build successfully during client generation.
• OpenAPI generation, documentation build, and repository lint pass.

[nicoloboschi]

I'd prefer to not expose the markitdown flag. maybe a better alternative is to expose if image (or some format) is supported or not as flag

for now I'll close until we have clear use case