Reworking and generalizing types.

[tobiasraabe]

• Date you used respy PyPackage: 06.07.2019
• respy version used, if any: > 1.2.1
• Python version, if any: 3.7
• Operating System: Windows

Closes #207, #202.

Current Behavior

See #207 for a discussion on the current state.

Todo

• Set target branch to develop once Implement parametrizations of Keane and Wolpin (1997). #209 is merged.
• Align the current covariates with KW97 by retaining regression tests.
• The predicted probabilities for each type should be called type_probabilities instead of type_shares.
• The covariates of the multinomial logit should be matched like choice covariates with parameters.
• Rename the coefficients of the multinomial logit from type_shares to type_prob. Also in optim_paras.
• Check _validate_params
• Random tests run with two or three type probability covariates.
• Delete _sort_type_info which will break regression tests.
• Sort type probability parameters so that the order in the frame is not important. Currently, we cannot implement that as _sort_type_info sorts types based on the first parameter which will change with sorting.
• Added test to ensure sorting of type probability parameters.

[tobiasraabe]

@janosg There are two issues left on the todo list, but they will break the regression tests because of type sorting. Maybe you can make a review now so that we keep the regression tests for your feedback. After that, I am going to implement the remaining issues and create 100 new regression tests.

[janosg]

Very nice PR. All the comments are minor.

[tobiasraabe]

@janosg Now, I will start breaking the implementation :).