Fix thumbnail and attachment multipart uploads

[aaronzi]

The thumbnail and attachment upload request bodies are mandatory, but their multipart schemas do not require the binary file property. An empty object therefore satisfies the OpenAPI contract even though there is no content to store.

The schemas also declare fileName as a separate property. In OpenAPI, each multipart schema property describes a distinct form part. RFC 7578 instead carries an optional source filename in the filename parameter of the binary part's Content-Disposition header. The abstract PutThumbnail and PutFileByPath operations accept only Part 2 File Content, which is defined as binary data and has no separate filename input. Part 1 File.value and thumbnail Resource.path remain model metadata and are not additional multipart parameters.

This change removes the redundant fileName property and requires file in every thumbnail and attachment upload schema. The file property documents the standard multipart filename mechanism, all affected service profiles and the combined API are kept aligned, and a dedicated artifact validator checks every multipart file-upload definition for the same contract. Because the incorrect schemas already existed in V3.1.3, the correction is recorded in the V3.2 changelog.

The AASX upload endpoints are handled separately in #656; this PR covers the remaining thumbnail and attachment uploads.

Validation

• python3 tools/validate_spec_artifacts.py
• python3 tools/validate_multipart_uploads.py
• 20 Query JSON Schema unit tests
• openapi-spec-validator passes for all six affected service profiles and the combined API
• Parsed all 21 multipart file-upload definitions and verified that file is required, fileName is absent, and the standard filename mechanism is documented
• git diff --check

Merge order

Depends on #662. Merge #646 through #662 before this PR.

[github-advanced-security]

QDJVMC found more than 20 potential problems in the proposed changes. Check the Files changed tab for more details.

[sebbader-sap]

I checked a bit whether or not PUT is a common practice for multpart operations: It's not. But I also didn't find anything which forbids it. Still, something to think about for the next breaking version.