change tracking update from nightly job to weekly

[FuhuXia]

Stats shows catalog visits to /_tracking in June doubles in number compared to April stats. The count of unique dataset page visits is four times more. Together with recent questionable Solr performance, they makes the nightly tracking update job takes too long to finish. Give the fact the page visit stats is not critical to be processes nightly, we can change to a weekly job.

Sketch

1. tracking-update is supposed to run nightly according to CKAN core. We need to change to default behavior in ckanext-geodatagov so it handles weekly data.
2. Change GH action cron to weekly.

[FuhuXia]

Stats were collected 3 days in April (10, 11, 12) and June (05, 06, 07).
Newrelic shows hits to /_tracking are 394K in April and 762K in June.
DB query shows 65K unique dataset visited in April and 256K in June.

--sample query
SELECT count(*) FROM tracking_summary
where package_id!='~~not~found~~'
and tracking_date in ('2023-06-05', '2023-06-06', '2023-06-07')

[Jin-Sun-tts]

catalog cron job change: GSA/catalog.data.gov#962

[nickumia-reisys]

Random follow-on notes: We suspect this issue is caused by bot tracking inflating the /tracking route. I thought a ticket was made, but apparently not. If we limit bot traffic to the catalog instance and this action should be able to be run nightly again.

[FuhuXia]

Random follow-on notes: We suspect this issue is caused by bot tracking inflating the /tracking route. I thought a ticket was made, but apparently not. If we limit bot traffic to the catalog instance and this action should be able to be run nightly again.

Not to limit bot traffic, but to exclude bot traffic from tracking count. We need to use above three days data, April (10, 11, 12) and June (05, 06, 07), combined with CloudWatch log data (containing user-agent info), to see the effect of excluding bot traffic to the tracking count.