Apply virtualization for large datasets.

[nicksimpkins]

If the dataset has 15,000 data points and your chart canvas is 1000px it is going to be rendering 5 data points within a single pixel. Not only does it take more time for the chart to load, but it is also hard for the user to see and hard for our crosshair to handle. I think we may do this to some degree already but it's been so long it is worth checking and likely updating.

[bilalaijaz12]

https://dev.to/crate/advanced-downsampling-with-the-lttb-algorithm-3okh