expand_corrmat about supereeg HOT 4 CLOSED

From (out of date) comments: Specifies whether to compute over all elecs (fit mode) or just new elecs (predict mode)

from supereeg.

I think we need to keep track of the signs of the entries of numerator:

• separately track positive and negative numerator values
• need to update model objects to accommodate this

from supereeg.

alternative idea: maybe we can set the numerator to complex numbers:

• positive values: np.real(numerator)
• negative values: np.imag(numerator)
• when adding, just add the corresponding real and imaginary parts
• when subtracting, swap the real and imaginary parts and then add

from supereeg.

create _to_log_complex(x): function that takes the log of the array x. positive entries in x will correspond to real parts and negative entries will correspond to complex parts.

also need a test for this function

from supereeg.