What problem does this solve or what need does it fill?

Compiling pipelines is expensive. Currently, we compile waiting pipelines in serial each frame.

I'm also not sure if the pipeline processing system can run in parallel with other systems.

What solution would you like?

Parallelize the inner loop over processing pipelines, and also the system if needed.

Other Context

I believe we're blocked on wpgu for this at the moment. Wgpu currently locks each object, so we can't really do things in parallel gfx-rs/wgpu#2710.