[Feature] Conditional spectral granger causality

[shimoura]

This PR is for the implementation of the Conditional Spectral Granger Causality. Currently, in Elephant, there is only the pairwise spectral Granger Causality.
This implementation is based on my own previous implementation reported here:

• Lima, V., Dellajustina, F. J., Shimoura, R. O., Girardi-Schappo, M., Kamiji, N. L., Pena, R. F. O., & Roque, A. C.. (2020). Granger causality in the frequency domain: derivation and applications. Revista Brasileira De Ensino De Física, 42, e20200007. https://doi.org/10.1590/1806-9126-RBEF-2020-0007

The theory should be equivalent to the following reference:

• Wen X, Rangarajan G, Ding M. Multivariate Granger causality: an estimation framework based on factorization of the spectral density matrix. Philos Trans A Math Phys Eng Sci. 2013 Jul 15;371(1997):20110610. doi: 10.1098/rsta.2011.0610. PMID: 23858479; PMCID: PMC3971884.

The current implementation is functional but the expected result is still not matching.

TODO list:

• double check method equations with theory
• add ground truth example (where?)
• add option of multiple trials. The multiple trials should be given to the multitaper method so it can be used to average the power spectra, then the Wilson factorization uses the cross density spectrum matrix from the already accounted multiple trials.

[coveralls]

coverage: 88.08% (-0.5%) from 88.605% — shimoura:feature/conditional-spectral-granger into NeuralEnsemble:master