fix(plugin-cloud-storage): prevent crop reprocessing during metadata persist

[kayareyouokay]

Summary

Fixes a regression in @payloadcms/plugin-cloud-storage where create-time uploads with crop edits could upload to S3 successfully but fail to persist the resulting document metadata.

Problem

When an image is uploaded during initial document creation with crop edits applied, the cloud storage afterChange hook performs a nested payload.update to persist returned upload metadata such as url.

That nested local API call was inheriting the original request's uploadEdits query state. Core upload handling then interpreted the metadata-only update as another crop/reupload operation. On initial create, the document did not yet have a usable persisted remote URL for that reprocessing path, so the metadata update failed and the server logged:

Failed to persist upload data for collection <slug> document <id>

This did not affect uploads without crop edits, or crop operations performed after the document had already been saved.

Regression likely introduced by #15570.

Solution

• remove req.query.uploadEdits before the internal metadata persistence payload.update
• restore the original uploadEdits value after the nested update completes
• keep the existing preserved file-buffer behavior intact
• add an integration regression test covering initial create with crop edits and asserting the document url is persisted

Testing

• Added an integration test in test/plugin-cloud-storage/int.spec.ts for the create-with-crop path.
• Passed locally with:
  PAYLOAD_DATABASE=sqlite COREPACK_HOME=/tmp/corepack pnpm exec vitest --project int test/plugin-cloud-storage/int.spec.ts -t "should persist upload metadata after an initial create with crop edits"