Handle regimes with only one average coherence value

[noah10]

Occasionally the sensor will not give us enough coherence values to generate an average coherence value (we need 480, or four minutes worth). It's possible that this can result in a regime with only one coherence value. When this happens we're unable to compute a 90%CI for the regime, because we can't calculate the standard deviation of a single value. In this case, we should treat the regime as having an 90% CI that overlaps the target average coherence (i.e., keep the regime in rotation). This will hopefully allow us to get one or more additional average coherence values for the regime so that the CI can be calculated properly.

This should be done for both the random and the resonance condition.

[noah10]

Note that in the unlikely event that a regime w/o any confidence intervals is the single "best" regime and the participant is in the resonance condition we will not split the regime and generate new ones, as we normally would if it had confidence intervals. The thought here is that we have so little data that it might well be "best" by chance and that the participant should practice it more so that we have a more representative average coherence for that regime.