Summary

Speeds up CI docs builds from a median 32m04s to 8m39s (3.7x).

• download up to four independent documentation prefixes from S3 concurrently
• customize HTML files with a process pool sized to the CPUs available to the job
• replace repeated full-document CSS selector scans with one consolidated document inspection
• reduce post-processing output to periodic progress updates while retaining warnings
• preserve the existing html5lib parser and byte-for-byte serialized output

Closes #448.

Validation

• verified optimized HTML output is byte-identical to the original serial implementation
• verified a second customization pass remains byte-idempotent
• verified the S3 scheduler never exceeds four concurrent prefixes
• verified a failed background S3 copy fails the overall script
• ran pre-commit run --all-files