Speed Improvements of Critical Collections

[doublebyte1]

This issue captures some ideas for speed improvements in access to feature collections, particularly REN and RAN.

• Disable the count parameter for those collections, as introduced in this PR. It should be noted that the server will no longer return the NumberMatched records.
• Fine tune the DB connection using the SQAlchemy driver options.
• Introduce the limit parameter at collection level. This will (hopefully) force the client to use paging? The combination of this and the disable of count seems dangerous.
• Create a cron job to restart the DB periodically, or (even better) after some indicators are triggered.
• Switch to use k8

In order to track this situation, we could explore these avenues:

• Create tests on GHC, to check the response times and trigger a warning when these increase.
• Create load tests on locust; compare k8 with the production infrastructure.

Before doing tests, it is important to identify the critical requests.

[PascalLike]

Similar to the PR above, there is another approach: this plugin https://github.com/cgs-earth/pygeoapi-plugins, specifically the PseudoPostgresSQL, uses database statistics to get the count, instead of performing the slow COUNT(*). This approach is not ideal if the data is updated frequently, because count is not calculated in real-time.