Vectorize significance testing, support more flexibility

[henrymj]

if significance_testing:
                p = np.zeros(len(self.t[self.t > 0]))
                # only test on timepoints after stimulus onset
                for i, t in enumerate(np.arange(len(self.t))[self.t > 0]):
                    # one-sided paired ttest
                    _, p[i] = sista.ttest_rel(
                        a=acc[t], b=acc_shuff[t], alternative='greater')
                # Use Benjamini-Hochberg procedure for multiple comparisons, defaults to FDR of .05
                _, corrected_p, _, _ = multipletests(p, method='fdr_bh')
                sig05 = corrected_p < .05

I think this should be pulled out into its own function that allows for specifying alternatives and correction methods.

I also want to note that scipy lets you test along an axis, and the following should work:

p = sista.ttest_rel(a=acc, b=acc_shuff, axis=0, alternative='greater').pvalue