SingleFile CLI 2.11.1

Nothing in the CLI itself changed. The release carries single-file-core 1.5.128.

Fixes from single-file-core 1.5.128

An archive saved with --crawl-save-archive-toc opens on its table of contents. The table of contents was written into the archive and could be reached by hash, but nothing ever went there, because the archive always opened on the first crawled page
A resource is identified by its bytes rather than by the content type the server sent. The bytes were read only when the header was missing or application/octet-stream, so a woff2 served as text/plain was embedded as data:text/plain. That costs twice: a browser with no format() hint in the @font-face rule has nothing left to identify the font by, and the archive writer deflates a file that is already Brotli compressed. The header is still kept when the bytes match nothing
--max-appended-data-length is honored by --crawl-save-archive. The options handed to the writer of a multi-page archive were copied by hand and that copy had dropped the budget, so it applied to a single page and was ignored for an archive of several. The list is now derived from the options the writer reads

AVIF, HEIC, JPEG XL and TIFF are recognized from their bytes
A video is identified as video/mp2t only when the sync byte appears at the 188 byte packet stride. Matching a single leading 0x47 was tolerable as a last resort behind a missing header, but it would relabel any video whose first byte is G now that the bytes win over the header

Co-authored by Claude (Claude Code)