Try to reclaim performance

[HighDiceRoller]

While 0.13.0 has a simpler interface than pre 0.13 the performance is much worse on some cases. I need to figure out what's going on and try to reclaim some of that performance.

• Caching?
• More expensive object startup?
• Core algorithm inefficiencies?

[HighDiceRoller]

The cached direction is quite a bit better; I've adjusted the algorithm selection code to favor it more. Still a factor of ~2x worse than pre 0.13.

[HighDiceRoller]

It looks like the dual-purpose == is hurting us here, checking logical die equality becomes a $n^2$ algorithm.

Options:

• Figure out some lazy evaluation scheme.
• Go back to only equals() for probabilistic equality.

[HighDiceRoller]

Lazy evaluation seems to work so far.

[HighDiceRoller]

Split off comparator handling to #48 .

[HighDiceRoller]

Seems good enough for now.