Issue with how weighted decile bins are created for distribution and difference tables

[Peter-Metz]

With the current code in utils.py weighted deciles cannot be created if too many filers have 0 or negative values for the relevant income measure (e.g. expanded income). The add_quantile_table_row_variable() function creates the decile bins in two steps:

1. After sorting the records by the relevant income measure, the records are broken up into 10 bins with equal number of people or filing units.
2. If decile_details=True, which is the default option for difference and distribution tables, the original deciles are further broken down to include a group of filers with negative income and a group with 0 income.

The problem arises if more than 10% of records have <=0 income since the bin sizes will no longer increase monotonically and pandas will throw an error when calling pd.cut().

This exact problem arose in #2444 when trying to run recipe05 with the new CPS. Since recipe05 replaces expanded income with "market income" (i.e. does not include benefits received), the number of filers with <=0 income grew bigger than 10% of the records.

A short term solution would include modifying recipe05. A more permanent solution would involve modifying the bin creation in utils.py.