General spatial panel optimization

[adkinsrs]

Currently working on optimizing the spatial panel for speed (with help from my AI buddies). There are some places where a callback triggers rerendering the entire plot, or panel object and I am trying to use some Plotly and Panel "update" functions to reuse the same objects. Also caching here and there.

[adkinsrs]

Closing this for now with recent code merges to devel. Currently things are snappy enough as is, but there are classes that have a lot of code duplication and could be refactored to inherit from a base class. When we create a multigene spatial display, it may be time to refactor then, as we are adding new routes to the codebase.

[adkinsrs]

Reopening this is a side issue I do in my spare time. Breaking this into a few subtasks

• Refactor SpatialPanel classes to use a common base class to drop duplicated code
• Drop plotly in favor of hvplot (high-level) or Holoviews (lower-level) if hvplot is not possible. Since they are in the same package suite as Panel and Datashader, they integrate much more nicely, thus becoming more performant and saving lines of code.
• (Lower Priority) Drop csv cache file support for parquet. The parquet format is still compatible with Pandas, but is more structured so we do not have to assign categorical fields as such after reading in the file (like I did with csv). When your file starts to get larger (~100MB), then parquet becomes more performant than csv. This is much lower priority has we have not hit those file sizes yet, and we can bridge the gap code by modifying existing csvs to Parquet or add "legacy" support. This bullet point can also apply to projection cached CSVs too.