Improve draw performance

[cvanelteren]

This PR introduces several improvements in drawing graphs. It achieves this by optimizing the draw request. Currently, when a draw requests comes in, the entire graph is redrawn, which can be inefficient. The new approach allows for more targeted updates, reducing unnecessary redraws and improving performance.

How it works When a draw requests comes in, we cache the layout to see if a layout needs to be adjusted (otherwise it is drawn immediately). If the layout did change, we reuse tick calculations and axes measurements, then we re-use cached layouts where possible to improve performance further.

The end results it that we can improve draw performance, especially versus tight layout in raw matplotlib.

EDIT: just wanted to emphasize that we are not optimizing how data is handled. So when the data increases the two performances will align. What we do here, is optimize the surrounding infrastructure to prevent unnecessary redraws when there is no change. I am trying to figure out if we can do the other stuff too but in a different PR.

[codecov]

Codecov Report

❌ Patch coverage is 90.28133% with 76 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
ultraplot/_animation.py	69.62%	24 Missing and 17 partials ⚠️
ultraplot/_layout.py	91.21%	16 Missing and 10 partials ⚠️
ultraplot/figure.py	89.83%	3 Missing and 3 partials ⚠️
ultraplot/axes/base.py	60.00%	1 Missing and 1 partial ⚠️
ultraplot/axes/cartesian.py	88.88%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!