Precompute picture noise instead of continuously computing it

[zshall]

As far as performance optimizations go, it would make sense to generate a few seconds of picture noise and repeat the same loop over and over instead of constantly churning out new noise patterns. I've noticed when doing performance profiles of TVS that picture noise calculation isn't great on lower-end devices.

[zshall]

Working on a solution that'll generate new picture noise as needed until a buffer is filled, then loop the buffer until changing channels. I'm thinking 180 frames of noise (3 seconds or so) should work well. This should be tested on a Raspberry Pi since I'd like to allow picture noise to comfortably run there.