RX syntax. Namely, a lot of code (used to) assume(s) that the
total-number-of-submatches (TSM) quantity is static even for dynamic
regexps.
Specifically, RE-TSM now returns an unspecific value instead of 0 for
non-regexps which should break most of the code that used to just
silently do the wrong thing. It's likely that more examples involving
dynamic sub-regexps will fail.
In the process, I also removed some of Olin's naming craziness (using
"%" signs to convey meaning) which I needed to do to halfway
understand what's going on.