feat: import data from the Publisher app database

[nacnudus]

Closes #586

The updated_at and public_updated_at fields in the Publishing API
(and downstream in the Content API) aren't good for telling when a
document was updated in the Publisher app, which is most mainstream
content.

Publisher app database backup files are now
available in a GCP bucket.

We're interested in ones with state 'archived' or 'published'.
Unfortunately the timestamps don't quite match the Publishing API
database, being out by a few seconds, so matching the editions will be
tricky.

An example query to join the Publisher data to the Publishing data, by
approximate timestamp:

with real_updates as (SELECT
  a.base_path,
  a.updated_at,
  a.public_updated_at,
FROM
  `govuk-knowledge-graph-dev.test.publishing-publisher` AS a -- the Publishing API data
where EXISTS(
  SELECT
    TRUE
  FROM
    `govuk-knowledge-graph-dev.test.publisher-updated-at` AS b -- the Publisher app data
  WHERE
    a.base_path = '/' || b.slug
    and a.updated_at >= b.updated_at
    AND timestamp_diff(a.updated_at, b.updated_at, second) <= 60)
),
latest as (select base_path, max(updated_at) as updated_at from real_updates group by base_path)
select * from latest order by updated_at

References:

• https://trello.com/c/pzNnETtk/119-add-data-about-when-a-content-item-had-either-a-major-or-minor-update-to-the-content-* api
• https://trello.com/c/J6zfW1EK/13-publishing-api-has-many-inaccurate-dates
• Present updated_at downstream publishing-api#1597
• https://gds.slack.com/archives/C02CM46TD52/p1688046251340989
• https://gov-uk.atlassian.net/wiki/spaces/CC/pages/32145674/Using+Publisher