feat: --scan-ext to scan for any file type, with multi-select

[xAlcahest]

Summary

--extract-scan only looked for streaming video. Now it scans for video and audio by default, and --scan-ext=iso,zip,mkv,flac (any extensions, any type) restricts the scan to those file types instead, which also covers a plain directory listing (Apache/nginx autoindex). The extension set is threaded through scan_page into classify_url and the URL-capture regexes are built dynamically from it. When a page yields more than one matching file, run_extract_scan opens a multi-select TUI (reusing tui_select_many) so the user picks which files go into the generated config; a deselected subset is captured as an alternation of the raw page refs so it still resolves on a relative-href listing. The download command flux <url> is unchanged and gains no scanning.

A directory-listing config is matched by its landing directory with an optional :port, so the generated config actually re-attaches to the page it came from.

Changes

• src/scan.c: classify_url takes the active extension set (default widened to video + audio + HLS, custom set matches only the given extensions, any type); dynamic bare-URL and href/src capture regexes built from that set and regcomp-verified; an extension-list series emits a list config keyed by extension; partial multi-select emits an alternation of the raw captured refs; reject candidates containing control characters.
• src/scan.h: extension-set params on scan_page; raw_ref on the candidate.
• src/text.c: --scan-ext flag (comma list, validated, rejects non-alphanumeric tokens and empty/trailing segments, only with --extract-scan); multi-select TUI in run_extract_scan; directory-derived match with optional port.
• src/test_scan.c: audio-default, --scan-ext directory-listing, partial-selection roundtrip, control-char, and ext-token validation cases.
• README.md, NEWS: document --scan-ext and the multi-select.