Shuffle null for session reliability metrics #174

Add p-values to the reliability metrics from #173. Each metric is recomputed --n-shuffles times after shifting every trial's trace by a random circular offset within its own epoch. p is the fraction of shuffles at or above the observed value, written to a <metric>_p column for each group.