Image range and histogram computation: cache, defer and/or optimize

[PaulHax]

To speed time to first image render, we can improve the range and histrogram computation in useAutoRangeValues:
https://github.com/Kitware/VolView/blob/main/src/composables/useWindowingConfigInitializer.ts#L31-L38

The image is not rendered until the autoRangeValues are computed. Can we defer the computation and do it in a web worker?

The above chunk of code takes ~570ms for the CT volume here:
https://volview-dev.kitware.app/?urls=[https://data.kitware.com/api/v1/file/63f39e907b0dfcc98f669cc6/download/joe-slim-pet-ct.zip]
Can we optimize/parallelize?

autoRangeValues is computed 3 times for the same image on the initial load. Probably not needed.

[floryst]

This is semi-related to streaming (i.e. computation on chunks and merging results) and making sure various pieces of the app (e.g. the transfer function widget) don't re-compute the range & histogram.

[floryst]

An implementation note: the linked PR merely shunts the computation to a webworker. It still computes the histogram once per view, which is still inefficient. This can probably be resolved via some store state that tracks requests for window/level histograms.

[floryst]

Ideally we want to compute the histogram once per image rather than once per viewport, but right now since the computation is not blocking it's not a major priority.