Chunk all the CUDA calls

[tbenthompson]

Currently, the screen will freeze if a user runs a large cutde calculation on the GPU that also drives their monitors. This is avoided by chunking the calculation in disp_blocks and disp_aca. Chunking allows a few processes to run in between the kernels and makes the screen much more responsive.

Ideally this chunking would be configurable since it seems to generally slow the computation by 5-10%.

[tbenthompson]

Alternatively, I suspect some asynchronous data transfer running simultaneously with the kernel might help make the chunking as fast as the unchunked version. The synchronous chunked version has lots of blocking operations.