Add Gradient Capability

[cjkunselman18]

This pull request uses Jansson derivatives to calculate and output the gradient of each contribution to the likelihood function for use in gradient-based optimization.

[nury12n]

I found that this part will also fail if one of the phases has a phase model only in sub-component space (ex. if you're trying to assess an A-B-C system, but one of the phase models is $(B,C)_2(C)$ with no A solubility) since the Workspace creation at line 320 takes in all the components.

My quick and dirty solution was to assume that parameters in sub-component spaces were fixed, so that the driving force gradients for these sub-component phases were 0. But a better solution should allow for assessing parameters in the sub-component spaces. A possible way would be to have additional PhaseRecordFactory objects for the sub-component spaces if needed and using the corresponding PhaseRecordFactory when creating the Workspace object, but not sure if this is most efficient solution.

Also, shifting the composition to be between 5e-6 and 1-5e-6 can be done with vertex_comp_estimate[counter] = np.clip(vertex_comp_estimate[counter], 5e-6, 1-5e-6)

[bocklund]

two changes I want to make and I think it's about ready to merge:

1. we copy of the Database objects in context creation to prevent mutating the variables in the global state to prevent weird behavior or race conditions with the lasting Database object
2. for performance, since gradients are a performance regression in function call time, i'd like to make the gradients behavior opt-in so that gradients are only computed if requested by the calling optimizer. this also has the benefit of allowing us to keep the test changes small and not thread through a bunch of API changes to handle gradient outputs