Linear algebra for 2x2 matrix

[tky823]

Numpy's algorithm of linear algebra is slow when a matrix is 2x2.
We can write down some algorithms naively.

[tky823]

np.linalg.inv for 2x2 matrix is slower than naive implementation.

[tky823]

np.linalg.cholesky for 2x2 matrix is NOT slower than naive implementation.

[tky823]

In "ISS2: An Extension of Iterative Source Steering Algorithm for Majorization-Minimization-Based Independent Vector Analysis, " a closed-form solution of the generalized eigenvalue problem for a 2x2 matrix is introduced, but this causes numerical instability in my experiments.

[tky823]

I will use the combination of np.linalg.eigh and np.linalg.cholesky for the generalized eigenvalue problem (for a 2x2 matrix). Naive implementation prevents numerical stability as described in #115 (comment).

Algorithm: https://netlib.org/lapack/lug/node54.html