Gadget interpolation speeds

[chrishavlin]

In playing around with the current json loader and datasets beyond IsolatedGalaxy, I've found that large sph particle datasets are pretty slow. e.g., the following json:

{"$schema": "https://raw.githubusercontent.com/data-exp-lab/yt-napari/main/src/yt_napari/schemas/yt-napari_0.0.1.json",
  "dataset": "snapshot_033",
  "field_type": "PartType0",
  "field_name": "density",
  "left_edge": [0.0, 0.0, 0.0],
  "right_edge": [25.0, 25.0, 25.0],
  "edge_units": "Mpc",
  "resolution": [600, 600, 600],
  "take_log": true}

takes around 2.5 minutes to load in napari. Most of this time is spent in yt, interpolating the SPH field. While alot of the slowness here may be unavoidable, there are at least three points that are worth looking at I think:

1. Is there a way to add some napari notification that the interpolation is in progress? The terminal log shows the usual yt notifications, but the napari GUI hangs while it waits for yt to return the interpolated data. Some sort of loading symbol would be nice...
2. There seems to be a significant differences in loading time between the yt code when executed in napari vs. executing the same yt code in a python shell: executing in napari was almost a whole minute slower. Need to do some more careful profiling to compare.
3. It might be worth considering a cache system -- the interpolated FRB would be fairly cheap to store on disk so that if you subsequently load an identical json it could reload the previously interpolated FRB. Could have an optional "use_cache" argument in json model. Not an immediate item to do, but may be worth considering in the future.

[chrishavlin]

For reference, the terminal log while loading:

[chrishavlin]

with the jupyter interaction from #22, I think this is less of an immediate issue, will remove from #7