Re: Use asv for running benchmarks

[Chris Warburton]

Made benchmarks for QuickSpec on TIP. We're reporting time taken,
precision, recall, #equations found, #conjectures possible and
#equations possible.

We also have a "benchmark" which just returns all of the generated data
(samples, QuickSpec output, analysis, etc) so asv will include it in the
output and we can commit it to git (for reproducibility, etc.)

Next steps are:

 - Do the same for hashspec and mlspec
 - Do the same for nat-simple, etc.

Then we can delete the convoluted mix of benchmark stuff we've got
littering the place.