Reduce run quantization overhead by generating lookup tables during bootstrapping, and using the tables for all subsequent run quantization.
Also rename run_quantize_*() to improve clarity. These tests demonstrate that run_quantize_ceil() is flawed.