fix: retry MediaWiki upload on 5xx instead of failing immediately

[DaxServer]

MediaWiki upload chunk/commit requests that return a 5xx (e.g. 503) were treated identically to permanent 4xx errors — the upload was marked failed on the very first attempt, with no BullMQ retry.

• Add MediaWikiServerError, thrown from apiUploadChunk when the upload endpoint returns a 5xx
• Rethrow it from the worker processor alongside HashLockError/StorageError/SourceCdnError so it escapes to BullMQ instead of being marked failed immediately
• This lets the existing attempts: 3 / exponential backoff retry policy apply to transient 5xx errors, same as it already does for source CDN 5xx errors

— Claude Sonnet 5

[greptile-apps]

Confidence Score: 5/5

This looks safe to merge.

• No blocking issues found in the changed code.

Important Files Changed

Filename	Overview
backend/src/mediawiki/client.ts	Separates retryable chunk upload server errors from terminal final commit server errors.
backend/src/workers/upload.worker.ts	Adds the new MediaWiki server error to the worker retry path.
backend/src/core/errors.ts	Adds the new MediaWiki-specific server error class.
backend/src/tests/uploadClient.test.ts	Covers chunk upload server errors and final commit server errors.
backend/src/tests/upload.worker.test.ts	Covers worker behavior for the new retryable MediaWiki error.

_{Reviews (2): Last reviewed commit: "fix: do not auto-retry the final commit ..." | Re-trigger Greptile}