Faster particles

[gcask]

On my system this yields a ~20% boost in FPS in Battlefield Huge.

The main tricks used here are buffer re-specification, which enables the code to avoid a cpu/gpu sync when the scene gets crowded as well as using a STREAM buffer (which is what it should be, allegedly).

It also uses less memory (now ~1.44MB, before ~2.36MB, ~40% less) and this in turn could help lower end devices as well.

Lastly it now uses the full range of a u16, instead of limiting the number of elements to a u8 making it draw even more in less.

[daid]

The SeriousProton RenderTarget render batching can most likely benefit from "Buffer_re-specification" as well?