Vectorize

[aditya-sengupta]

• The addition of a "vectorized" mode for wavelets.compute_wavelet(), so that the transform does not loop over all omegas and taus, i.e. the transform ideally takes less than O((num_omegas) * (num_taus))
• Some numpy optimization
• Expanded basic test, added data/test_vectorized_timing.py

Still a WIP, current issues/enhancements:

• vectorized mode passes basic test, but causes S matrices to be singular (unclear why) in larger-scale test: to view the problem, run data/test_vectorized_timing.py
• some substeps are still unvectorized and done naively in list comprehensions
• vectorized mode does not call subfunctions: ideally after the above steps are done, the intermediate functions should be rewritten to work either on one (omega, tau) at a time or on all at once.

[tzdwi]

This looks awesome; I'm a little short on free time the next few weeks, but let me know once you're happy with everything and I'll test it all out!