feat(extract): auto-detect next disc, drop double confirmation

[Xitee1]

Summary

• Discs ≥ 2 are detected automatically by polling drive_status (CDROM ioctl). No more prompt_yn(\"Insert another disc?\") + prompt_disc(\"Insert disc N\") per transition.
• Single-keypress e (via termios cbreak on stdin, no Enter needed) signals "done collecting, proceed to extract" — replaces the trailing Y/n confirmation.
• Disc 1 keeps the classic Press Enter when ready prompt so the user can read the header info before the run starts.

Implementation

• New src/bd_archive/ui/keypress.py: cbreak_stdin() context manager (saves/restores tcgetattr, no-op on non-TTY) + read_keypress(timeout) (select-based, lowercased single char or None).
• New _wait_for_next_disc(dio, mount_dir, target) in commands/extract.py: polls drive + stdin every 300 ms, renders a Braille spinner on TTY stdout. Returns mount path on disc detect, None on e.
• Wrong-chain / unparseable disc → existing continue path runs the finally (umount + eject) and the next iteration re-enters the wait. No new validation logic.
• _mount_with_prompt stays untouched — still used for disc 1 and the par2-damage re-insert path (same-disc semantics, different concern).

Test plan

• Normal 2+ disc chain: disc 1 prompts, disc 2 auto-detected after eject + reinsert.
• Press 'e' after disc 1 → extraction runs against just that disc.
• Insert a wrong-chain disc as disc 2 → warning + auto-eject + resumes polling.
• Insert a blank / non-archive disc → warning + auto-eject + resumes polling.
• Ctrl+C during the wait → clean exit 130, terminal not stuck in cbreak.
• bd-archive extract … | tee log.txt (non-TTY stdout) → no spinner spam, still works.
• bd-archive extract … < /dev/null (non-TTY stdin) → warns about 'e', polling still functions.

🤖 Generated with Claude Code