Updated temperament search code

I've updated these files:

http://x31eq.com/temper/regular.py

http://x31eq.com/temper/regular_wedgie.py

Firstly there was a bug in the TOP optimization where it was calling the library on every loop over the primes, instead of only at the end. So that means it should only have been slow, instead of horribly slow.

I've also generalized the code for producing equal temperaments to order, so that TOP is on an equal footing with the weighted prime RMS. The result is that the TOP test only takes 10 times as long as the equivalent RMS test! Progress indeed!!! The RMS does take a bit longer than it used to, though, because the new code generates tens of thousands of classes to handle the generality.

Graham