Python scripts to BYM2 case study

[farhanreynaldo]

The following PR is a draft on Fitting the Model to the Scotland Lip Cancer Dataset and Bigger data: from 56 counties in Scotland to 1921 census tracts in New York City sections.

There are several issues need to be addressed.

• The BYM2 model cannot be directly translated from R to Python because as far as I know, there is no equivalent INLA library in Python.
• The raw number of 2001 events plot has different value scale compared to the R implementation. The Python plot has a range of value from 0 to 300, while the R plot has 0 to 15.
• There is an exception for BYM model in Python, however the summary still produced.

[bob-carpenter]

Pinging @mitzimorris, who wrote the original ICAR case study.

If you want a dependency on a low-profile package, there's https://github.com/al-jshen/pyinla, but it looks like it just wraps the R version, so probably not a good idea.