Change PostgresSearchBackend to optmize for UUID primary keys

[thiagoferreiraw]

Context

After running some performance tests, I realized the way we are casting types for the model primary key vs searchentry.object_id was unefficient.

When we do a model.uuid_pk::text conversion, postgres lose the ability to use the primary key index and that makes the query way slower.

Queries and plans (testing with 100k records)

Note the primary key index was not used because of a type cast (partnershipledger.id::text)

In nearly 100k records, the execution went from 71ms to 7ms 💯

Before

After

Proposed sollution

If the primary key is UUID, we cast type in the watson_searchentry.id instead.

[thiagoferreiraw]

Hi @etianen!

Here's a performance improvement we made recently in our own codebase and we'd like to extend that to the community.

Let me know your thoughts about the proposed sollution and/or if I need to adequate the code to meet the project standards.

Thank you!

[etianen]

This is a great change, thank you!