Decouple spline staging from class declaration

[mlincett]

Addressing #191 as intermediate step for #189.

This PR decouples the spline staging and instantation of the photospline services from the class declaration, so that importing the class does not trigger file writes or memory allocation by I3PhotoSplineService.

I have the kept the structure of having everything implemented in static attributes / methods, so the class is still not meant to be instantiated, but I am considering taking a step further and changing this as well.

In principle, aside from readability concerns, there is nothing preventing us to define the reconstruction traysegment as a bound method (so I have been told), so instead of:

tray.add(RecoAlgo.traysegment)

one would have:

reco = RecoAlgo(...)
tray.add(reco.traysegment, ...)

I think this would provide more flexibility as well as a natural way of implementing configurability for the reconstruction algorithm (see #188).

[tianluyuan]

In general I'm in favor of not loading in data files until it's needed. I think this would cause issues for parallelization as in #140, though I'm not sure if that is still being pursued.

Perhaps @kjmeagher and @briedel can comment, or at least take note of this PR.

[ric-evans]

In general I'm in favor of not loading in data files until it's needed. I think this would cause issues for parallelization as in #140, though I'm not sure if that is still being pursued.

Perhaps @kjmeagher and @briedel can comment, or at least take note of this PR.

Ah yes, parallelization 😄 ewms-pilot enabled on-worker parallelization (called "multitasking") in v0.9.0 (Observation-Management-Service/ewms-pilot#28). Now that #174 has been merged, multitasking can be enabled.

Using this + multitasking + #174's updates helps to avoid the memory concerns that were addressed in #140. Of course, we'll have to test this to see what else needs to be addressed, but it's generally using a different paradigm.

I'm planning to begin work on this in June (after vacation 🌴)

[kjmeagher]

This is fine with me. This is basically the same thing as I suggested

[ric-evans]

new issue: #194

[ric-evans]

Good improvements. I have a couple requests, mainly for coding style.

[mlincett]

There are no blocking concerns and I have manually tested millipede_wilks manually. I am going to merge.

We can then move the discussion to #189 .