More testing/example data #13

These kinds of testing data are currently available:

• Sentences randomly generated by hypothesis
• CNF sentences from SATLIB benchmark problems
• d-DNNF sentences compiled from the previous sentence with D#
• sd-DNNF sentences expressing linear orders, from examples/socialchoice.py

It would be good for benchmarks and such to have sentences that are useful (as in having some application outside this library), structurally complex, and hard to analyze.

Sentences that are not in NNF could perhaps be converted to NNF with nnf.operators.