Added SLSQP minimizer and fixed unittests

[nightvision04]

• Boosted indicator speed by 4-10x
• Boosted lppls.minimize() speeds by at least 25x
• Fixed unittests

[nightvision04]

@Joshwani Just so you know, I tested this change for stability between datasets. I flagged the same zones after changing the minimizer function. It is so blazing fast. Using it on my local branch and it really is incredible. Showed me where the new speed bottlenecks are to work on.

[nightvision04]

I'm glad I wrote unittests beforehand, haha!

[nightvision04]

@Joshwani

I've noticed that the SLSQP method, on average requires far fewer searches than Nelder-Mead.

This has me a bit concerned, because it means that "condition 1" will be flagged far more often with SLSQP.

Since SLSQP agrees with Nelder-Mead's solution (when found), I am not too concerned that we are failing to meet assumptions about the data (convexity, nonlinearity, etc).

Was there a particular reason that you have been using Nelder-Mead?

[nightvision04]

Also - this branch shouldn't produce any merge conflicts with the numba branch, in case you are concerned.

[Joshwani]

This is great stuff @nightvision04!