TODO

These are the things I'll get to do when I get to do them:

  * blow-ups, tuning of IMG
  * adding machines of fractional (e.g. 1/7) amplitude, for more tunings & matings
  * "dilute" FR machine by multiplying stateset with {1..n} and chaining them in sequence
  * actual matings
  * given obstruction, compute quotient machine. More generally, find a common format for obstructions
  * canonical decompositions
  * recode properly rpoly, maybe have higher precision?

  * "prenucleus" using just the states, not the action to compare. possibly return also the relations that are needed to make it a nucleus, something like the "triangles".
  * abstract triangulations: just work with pairs/triples of conjugacy classes. do the lifting symbolically. train tracks.
  * submachines, esp. with subalphabets