support very large stimulus sets

[neuromusic]

currently, there are two options for stimulus generation for behavior: "block" and "random". in both cases, a static stim file is read and the full stimulus set is read in and presented either in order (block) or shuffled, then presented (random). this is not optimal for large stimulus sets (over 1000 stimuli), as they take a very long time to load and impossible for very large stimulus sets (over 2000 stimuli).

some experiments may generate a large number of unique stimuli based upon a small number of experimental parameters (e.g. probabilistic sequence discriminations, psychophysical discriminations).

solutions:

1. deal with it. generate blocks externally & only generate blocks < MAXSTIMS (however, fixing fix stim code collision for large stimulus sets #11 will be important)
2. add a new "static block" stimulus selection protocol. rather than reading in an entire file, this protocol would simply hold onto the file handler, then read each line at the beginning of a trial. the block would terminate when we hit the end of the line