fix: paginate OSS deployment object verification

[JJasonSun]

Summary

Deployment verification reads every OSS object-list page so prefixes containing more than 1,000 objects do not produce false missing-object failures. It follows NextMarker until IsTruncated is false, handles XML with or without a namespace, and rejects missing or non-advancing pagination markers.

Related: #58, #59 and the failed deployment verification.

The pagination behavior follows the OSS ListObjects contract.

API contract

Not changed.

Validation

• pnpm run validate passed locally with Node 26.7.0, pnpm 8.10.2, and Python 3.12.14: type checks, production build, distribution verification, 13 deployment tests, and 22 browser tests.
• Regression coverage includes 1,001 objects across two pages, namespaced and plain XML, and invalid pagination responses.
• Read-only public verification matched all 386 objects from the latest deployment artifact. This confirms the existing site content, not execution of this new pagination code against production OSS. No production deployment was run for this change.

Checklist

• I kept the change focused and public-facing.
• I removed secrets, private URLs, phone numbers, and customer data.
• Any OpenAPI change describes behavior supported by the public API. (Not applicable.)
• I updated tests when observable site behavior changed.
• I added screenshots for visual changes. (Not applicable.)
• I ran pnpm run validate.

[JJasonSun]

Production deployment succeeded on September 15, 2026: run 34939745026. The new pagination code listed 1,063 objects and verified all 386 uploaded objects without a missing-object error. The workflow's public-file byte comparisons also passed.

An additional read-only public audit matched 385 of 386 objects byte-for-byte. The remaining cached debug source map differs only in the build timestamp in its source URL; its source content, mappings, and the corresponding runtime JavaScript match. This does not affect the deployed pagination fix.