Enable additional attributes with CounterfactualGenerator, improve robustness

[dylanbouchard]

Is your feature request related to a problem? Please describe.
Currently, CounterfactualGenerator offers functionality for two attributes, race and gender. While these are important attributes for assessing fairness, other attributes are important as well. Attributes to consider adding include, but are not limited to, sexual orientation, disability/able-bodied, age, and religion.

Additionally, the current word lists for race and gender may not be exhaustive, so improving the coverage of these word lists would improve the performance of this technique.

Describe the solution you'd like
The solution should enable additional protected attributes for counterfactual substitution and/or improve the robustness of the functionality for existing attributes.

Additional context
Source code to be modified includes langfair/generator/counterfactual.py and langfair/constants/word_lists.py