Here's a revised proposal for the timed lock algorithms for multiple
lockables.

Changes:
* They now return `int` and the best motivation I could think of
   for that was added to the rationale.
* My reason for why they accept zero or more lockables was added
   to the rationale.
* "sentence 11" eradicated.
* "instroduced" corrected.
* Typeface for "unlock" corrected in sentence 7.
* [[nodiscard]] discarded.

* Reference implementation changes:
   * Updated to return int.
   * Now yields to be both smart and polite.
   * Removed residue from an earlier version in which there was a loop
     around `try_lock_until()` to retry when it fails spuriously. If
     try_lock_until() now fails before the timeout, so do the
     algorithms. Is this something I should rethink?

I'm guessing these changes don't need to go into the paper before it
gets an official number?

Best regards,
Ted Lyngmo