Filter / separate out CI downloads where possible

[brynpickering]

PyPI download stats are heavily skewed by CI action downloads. Their bigquery table does distinguish between these downloaded sources. E.g., a query that filters out CI downloads:

SELECT
  COUNT(*) AS num_downloads,
  DATE_TRUNC(DATE(timestamp), MONTH) AS `month`,
  file.project AS `project`
FROM `bigquery-public-data.pypi.file_downloads`
WHERE
  -- Only user downloads, not downloads as part of CI pipelines
  details.ci is NULL
  -- Only query the last 6 months of history
  AND DATE(timestamp)
    BETWEEN DATE_TRUNC(DATE_SUB(CURRENT_DATE(), INTERVAL 1 YEAR), MONTH)
    AND CURRENT_DATE()
GROUP BY `month`, `project`
ORDER BY `month` DESC

It gets expensive to query the table quite quickly, sadly.

The same filtering is possible with the Julia package stats table, using the "client_type" columns.