Copulae

Enable users to pass in uniform marginals straightaway

1. Fixed teething issues in the Empirical Copula
2. Minor changes to the psuedo_obs function

1. Refactored estimation codebase
2. Fixed bug when fitting Archimedean copulas without adding initial paramter (Issue 29)
3. Add python 3.9

1. Added MarginalCopula. This enables users to specify different marginal distributions given a single dependency structure
2. Fixed bug in EmpiricalCopula caused by eager conversion of data to pseudo-observations
3. Update EllipticalCopula getitem and setitem logic and removed delitem to better handle parameter updates
4. Convert BaseCopula to a Protocol typing
5. Introduced 3.8 typings and made it backwards compatible by requiring older versions of python to install typing extensions

This release brings in the following changes

1. The addition of the Gaussian Mixture Copula Model (GMCM).
2. Backport the code to support Python 3.6
3. A modest refactoring of the copula's Estimator class
4. A modest refactoring to combine the AbstractCopula and BaseCopula class while retaining type hintings (for PyCharm)