Mobile performance

Topic

As p5 sees increased adoption for mobile AR/VR, what can we do to make sure things run more smoothly? Here are a few points to discuss:

• After https://github.com/processing/p5.js/pull/5802, stroke rendering is a bit heavier than it used to be. Is this a bottleneck? Should we add a NONE stroke cap/join type so that applications targeting mobile can opt out of the extra geometry of higher fidelity strokes?
• After https://github.com/processing/p5.js/pull/5915, per-stroke colors are used to allow interpolation in the same way fill interpolation works. Is this a bottleneck? Should we offer methods to disable all per-vertex colors for the sake of performance?
• Drawing the same shape repeatedly would be more efficient if it is cached on the GPU as a p5.Geometry. Is this an optimization that most sketches would benefit from? If so, should we provide an easier way to create these? (https://github.com/processing/p5.js/issues/5393)

In general, it seems like we should do some profiles to get some data on what parts are slowest on mobile in order to decide what approach would help.

Review PRs 5802 and 5915 to understand the stroke geometry and per-stroke color changes, then read issue 5393 about caching repeated shapes as p5.Geometry. Profile representative mobile AR/VR sketches to identify the slowest areas; done means the profiling data supports a clearly scoped optimization direction.