Fix issue with concurrent identical searches (Fixes #233), Change "log" level messaging to "debug1" (Fixes #242)

[bitner]

Fixed

• Change all "RAISE LOG" instances to "RAISE DEBUG1" (Fixes Excessive growth of database logs #242)
  • In client_min_messages, "LOG" level is lower than "NOTICE", but in log_min_messages "LOG" level is higher leading to huge growth in messages going to the log even though they don't show up with client_min_messages. This changes the use of "RAISE LOG" to "RAISE DEBUG1" which is lower than "NOTICE" in both client_min_messages and log_min_messages to allow for multiple levels of verbosity.
• Refactor search_query and search_where functions to eliminate race condition when running identical new searches concurrently. (Fixes Occasional syntax errors from well formed search queries #233)
  • Reorders how we check if we need to run stats so that we can return as early as possible when we know we don't need to update statistics.
  • Does not use cache for where and order by statements any more. These functions are quite fast and there are no mechanisms to clear cached values when any changes in search to SQL function code changes.
  • Improve timing of using SKIP LOCKED to allow for better ability to reduce lock contention by skipping updates to lastused and usecount fields under concurrent identical queries.
  • Fix some issues with read only mode to make sure we don't try to update statistics when using read only mode.

[bitner]

@drnextgis or @mrob95 can one of you try to get the race condition using this branch? I haven't been able to replicate it.

[drnextgis]

I've tested it using this branch and no longer encountered the problem.

[drnextgis]

Glad to see it's been merged. Looking forward to the new release!