Logbook speedup

[amelchio]

Description:

The logbook often turns event data into a State object twice before dropping an event.

This PR optimizes that down to zero State constructions for discarded events:

• Some filters are moved earlier, before the final State construction
• All filters are rewritten to work directly on event data, removing the first State construction completely

This gives me a solid 45% reduction on logbook loading so I wanted to get it out for comments even if I have not tested it much yet.

Credits to @OverloadUT for pointing me in this direction.

Related issue (if applicable): fixes #

Checklist:

• The code change is tested and works locally.

If the code does not interact with devices:

• Local tests with tox run successfully.
• Tests have been added to verify that the new code works.

[balloob]

Creating a tuple for each event is also slow. We should move this out of the for loop.

[balloob]

It's confusing that this var name is the same as the one used above.

Will also mess with if statement below??

[amelchio]

It is the same variable. We look for domain and entity_id in each kind of event and do the include/exclude filtering if one is found.

[balloob]

We should migrate the filters below to helpers/entity_filter.py, as that one returns an optimized function for the passed in configuration.

[amelchio]

All the excluded_domains etc. filters? Agreed, though that seems like a separate opportunity that could fit in a separate PR :)

[balloob]

So if you want to be real scientific, and actually come up with measurements to proof your awesome work, consider writing a benchmark. (optional)

Home Assistant benchmarks

[amelchio]

I will try to add a benchmark tonight.

[amelchio]

Added a benchmark that is 35% faster for events that we keep and 90% faster for dropped events. I don't think we should treat this as too scientific but it definitely seems worth the trouble.