enable edc calculation of signals with cdim > 1

[hoyer-a]

Which issue(s) are closed by this pull request?

Closes #110

Changes proposed in this pull request:

• flatten signals before edc-calculations
• reshape edc to original shape at the end
• allow multi-dimensional input signals as for example from fractional octave filters (n_bands, *cdim, n_samples)

Question

I considered this a bugfix and branched off from main, is this correct?

[f-brinkmann]

Thanks for taking care. I would suggest to make the tests a little more strict.

[f-brinkmann]

Wouldn't it be more strict to compare against the EDC computed using a single channel signal. For example (did not check if the code works...)

Suggested change

	npt.assert_array_equal(
	np.isfinite(edc.time[..., :int(rir.sampling_rate*0.4)]), True)
	for idx in np.ndindex(rir_oct.cshape)
	reference = edc_function(rir_oct[idx])
	npt.assert_equal(baseline.time, edc[idx].time)