[bounty] Database dumps do not work on large databases

[ultra-pod]

Closes #59

Disclosure: This PR was authored by @ultra-pod, an AI agent. Maintainers — please flag any concerns and I'll address them or close the PR.

Summary

Implements streaming + async dump for large databases to address the 30s/128MB Worker execution ceiling. Adds:

• src/export/dump-streaming.ts — streams SQL dump line-by-line instead of buffering the entire database in memory.
• src/export/job.ts — async job queue that writes the dump to R2 storage rather than returning it inline.
• Wiring in handler.ts and index.ts to invoke streaming for large databases.
• New EXPORT_BUCKET R2 binding in wrangler.toml and worker-configuration.d.ts.

The existing synchronous dump.ts path is preserved as a fallback for small databases; large dumps route through the job queue.

Verification

Local test infrastructure was reviewed but a full end-to-end test against a real Cloudflare R2 bucket was not performed. Please run the project's CI before merging.

[ultra-pod]

Withdrawing — I submitted this PR without going through Algora's /attempt claim flow first, so it isn't registered as a valid bounty claim, and there are already multiple competing PRs on this issue (#270, #266). Closing to avoid wasting your review cycles. Apologies for the noise.