CI build reliability: spurious failures blocking PRs

[goelakash]

Several PRs are currently blocked by CI failures unrelated to their changes:

1. x86_64-tensorflow has a 15-minute timeout but mamba dependency resolution alone takes longer than that on uncached PR builds. All other tensorflow variants get 25 minutes.
2. Docs linkcheck fails on external sites that are down or have moved (unixodbc.org unreachable, opensource.org and docs.docker.com redirects treated as errors).

Open PRs to address these:

• Bump x86_64-tensorflow timeout to 25 minutes
• Ignore unixodbc.org link failure
• Ignore huggingface.co rate-limit failures (Ignore huggingface.co in sphinx linkcheck #2475)
• Bump build-docs timeout to 15 minutes (Bump build-docs timeout to 15 minutes #2476)
• Fix docs.docker.com redirect
• Fix opensource.org redirect

[goelakash]

Related PRs addressing these issues:

• Increase x86_64-tensorflow build timeout to 25 minutes #2465
• Add development setup guide for local pre-PR checks #2461

[goelakash]

Need more logs to understand why we fail with timeout for pip installs