performance optimizations #1017

Doing some performance optimizations in the ndwidget branch.

• lighting objects are only present in a PlotArea.scene when a mesh is added to it. This buys us a lot of fps when using lots of subplots, went from 40 -> 50 with an NDWidget viz with 7 subplots.
• optimize Axes update. I added an Axes._get_state() which returns a tuple[camera_matrix, viewport_rect, viewport_logical_size, scale]. The axes update only if this state changes. Went from 50 -> ~65fps !
• we're probably doing a bunch of copies all over the place, use astype(copy=False) when necessary
• implement yuv and other colorspaces and "bufferless" TextureArray #1033
• disable AA by default, pixel_scale = 1.0 by default, top UI menu to set renderer options
• Reactive rendering #1018

Here's a flamegraph with pyinstrument on some NDWidget stuff with 7 subplots:

I think the only way to bump up the performance more (other than IO and NDWidget-specific stuff) is reactive rendering, I will make a dedicate issue about that.