Update demographics.py to handle data downloaded with forecasted values

[SeaCelo]

Added code to filter by forecast model variant. This is necessary as described by issue #37 . If the user specifies a data range that includes data generated by a forecasted model (currently, this means any year >2021), the returned data will have multiple model variants for each observation. This needs to be filtered. We default to only using the "median" forecast model.

Added adjustments to the following functions: get_un_fert_data(), get_un_mort_data(), get_un_pop_data().
The adjustments:

• create an additional parameter called variant that defaults to "4" (median)
• Adds a description of this parameter
• Filters the results to only include observations where the variable VariantId matches the parameter variant

[jdebacker]

@SeaCelo This is a nice addition. However, with PR #51, we'll move over to using the demographics.py module from OG-Core. Could you make this contribution to that project?

[SeaCelo]

This PR is no longer necessary. Will make any contributions to the new demographics.py in OG-CORE. Closing.