Imported from https://youtrack.roxen.com/issue/PIKE-53

Reported by Martin Karlgren marty@roxen.com

Search.Database.MySQL: get_document_id and get_uri_id make SQL queries for each lookup, and it should be possible to cache that in memory. Negative caching is needed too. Cache entries need to be invalidated appropriately when any remove* method is called.

Also add a method that accepts an array(string) of URI:s to populate the two caches for, which should then read entries batch-wise from the database.

The purpose of these improvements is primarily to improve batch performance, i.e. re-indexing.