Fully expose parallelism in API

[jdwhite48]

Using parallelism seems to be at worst comparable to sequential for the performance of Groth-Sahai. However, I imagine there are use cases where sequential computation would be more desirable. And so, to fully expose parallelism in the API, a number of changes would need to be made, offering parallelized variants for the following algorithms for field, B1, and B2 matrices:

• scalar_mul
• transpose
• add
• par_*mul

And the API should also expose the option for parallelism

• Also offer par_commit* variants which use the aforementioned parallelized computations (macro'ed?)
• Also offer par_prove* and par_verify* variants which use the aforementioned parallelized computations (macro'ed?)