Merge missing data encodings

[rolyp]

See notebook. Summary:

• read dataset using utils.read_data
• examine some values in LRE Ages 3-5 - Full Incl # column
• plot the frequences of the unique values in the subsample
• instantiate Ptype and fit schema to subsample
• plot posterior distribution for column type and row type
• list the missing values for the column
• replace those values in the column by a new missing data encoding
• run PType again to verify new encoding correctly identified as missing

To do:

• read dataset directly rather than via utils.read_data
• nothing gained by plots – remove
• use Ptype to browse unique values?
• subsume missing probabilities plot with col.get_missing_values()?

[rolyp]

@tahaceritli Do we actually need this use case? Isn’t merging of missing data encodings exactly what is achieved when you do Schema.transform, when all data values interpreted as “missing” are mapped to pd.NA?