Disaggregation stats in Open SDG output

[brockfanning]

This adds an "endpoint" at /stats/disaggregation.json to the Open SDG output. This includes the following data:

• disaggregated: Number of indicators that are fully disaggregated
• not_disaggregated: Number of indicators that are not fully disaggregated
• total: Number of statistical indicators (this will be equal to the total of "disaggregated" and "not_disaggregated")
• disaggregated_percent: Percentage of indicators that are fully disaggregated (percentage of "total")
• not_disaggregated_percent: Percentage of indicators that are not fully disaggregated (percentage of "total")

The way that it determines "fully disaggregated" is dependent on a metadata field called expected_disaggregations, which should be a list of column names.

For example, if an indicator is expected to be disaggregated by "Age" and "Sex", then this metadata field should be something like (in YAML):

expected_disaggregations:
  - Age
  - Sex

If the indicator's data itself has both an Age and a Sex column, then it is considered "fully disaggregated".

If an indicator does not have an expected_disaggregations metadata field, then it will automatically be considered fully disaggregated.

The next step is to add code to Open SDG to use this data.

[LucyGwilliamAdmin]

I started testing this and then realised it's still marked as draft, but leaving a comment here about my error: https://github.com/ONSdigital/sdg-data/runs/1365907679?check_suite_focus=true