Merge validate_spatial and validate_series

[intarga]

No description provided.

[Lun4m]

It's going to take a while 😅 In the meantime, I have a question regarding a doubt I've had since we merged the two caches: how expensive is it to build the R*-tree? We are building it every time we call fetch_data and maybe some pipelines don't even need to run spatial tests, so I was wondering if inside DataCache we should instead have an Option<SpatialTree>? And maybe a constructor that leverages the new SpaceSpec? Or it doesn't matter?

[intarga]

how expensive is it to build the R*-tree?

The algorithm is n * log(n), so I don't think it's a concern.

I do agree there may be a need to optimise the case of simple pipelines though, as the fresh pipelines will likely form the vast bulk of our request load. I think if we're going to push to optimise those though, we should be more aggressive in our optimisation strategy. I'm thinking a much simpler DataCache that removed the spatial elements and the outer vec entirely, along with a cache of the recent data (a hashmap keyed by timeseries id, containing ringbuffers of data) bypassing the need to query the DB and do a network call entirely. That was the idea with #77.

It would be good to see how the more general validate function performs with real pipelines, a representative data flow, and a real DB connector though, so we can see how much optimisation it needs, and have a target to aim for.

[Lun4m]

Looks good, I only have minor comments