Include ISO 3166-2 subdivisions in client region field annotation

[critzo]

The annotation service currently populates connection_spec.client_geolocation.region with the top level ISO 3166-2 region code. To restore the region code granularity we had prior to 2017-05-11, we should add a field for the client region subdivision, annotated with the subdivision codes in the ISO 3166-2 standard.

To demonstrate the issue, I include the query below, and a trimmed result set for Great Britain. Prior to 2017-05-11 we annotated ~198 region codes in this field, whereas now it's 4:

SELECT partition_date, connection_spec.client_geolocation.region FROM `measurement-lab.release.ndt_all`
WHERE connection_spec.client_geolocation.country_code = 'GB'
AND partition_date BETWEEN '2017-05-10' AND '2017-05-12'
GROUP BY partition_date, connection_spec.client_geolocation.region
ORDER BY partition_date, connection_spec.client_geolocation.region

# Result:
partition_date	region
2017-05-10	A1
2017-05-10	A2
2017-05-10	A3
# + 195 more FIPS 10-4 regions in Great Brittain
...
2017-05-11	England
2017-05-11	Northern Ireland
2017-05-11	Scotland
2017-05-11	Wales

[critzo]

Noting that the ISO 3166-2 region codes are a part of the Maxmind Geolite2-City Locations.

The fields I believe we should consider adding to the ndt schema and annotator are:

• subdivision_2_iso_code
• subdivision_2_name

[critzo]

Adding a link to the place in the code where the current region code is referenced: https://github.com/m-lab/annotation-service/blob/master/geolite2v2/geo-ip.go#L111

Should be straightforward to pull the ISO 3166-2 subregion 2 code from the City dataset for the sub-regions.

[stephen-soltesz]

m-lab/etl-gardener#281