Position-only vertex buffer for shadow map / depth prepass

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

Non-position vertex attributes are not necessary for depth-only views like shadow map views or the depth prepass.

Separating vertex buffers into one for positions only, and one for everything else, _may_ improve cache performance for depth-only view rendering.

We'll have to see if it's worth it.

## What solution would you like?

Have one vertex buffer for positions, and a second vertex buffer for everything else.

The issue names no files, tests, or entry points, so first locate the vertex-buffer and shadow-map/depth-prepass rendering paths in the Bevy repository. Determine whether position-only buffers improve depth-only rendering, benchmark the relevant views, and consider the work done when the result is supported by measurements and the two-buffer arrangement is implemented if worthwhile.

Written by the indexing model from the issue text.