gh-228: results with multiple \ell axes

[ntessore]

Adds support for multiple $\ell$ axes to the Result class. This can be used to read and write, e.g., covariance matrices.

I had to add this to Result instead of the CovMatrix class since I couldn't figure out a good way to recover the correct class (Result, CovMatrix, or potentially classes for spectra or mixing matrices) from either read() or binned(). With one multi-purpose class, the problem doesn't arise.

To create a covariance matrix or other result with a second $\ell$ axis:

cov = heracles.Result(
    arr,
    ell=(ell_1, ell_2),  # optional
    axis=(-2, -1),  # optional
    lower=(lower_1, lower_2),  # optional
    upper=(upper_1, upper_2),  # optional
    weight=(weight_1, weight_2),  # optional
)

What determines the $\ell$-dimension of the result is the value of result.axis. If axis is not given explicitly, the last len(ell) axes are used (e.g. -2, -1). So either ell or axis must be given for multiple $\ell$ axes, otherwise, a single $\ell$ axis of -1 is assumed.

Closes: #228

[JaimeRZP]

LGTM except for the fact that I understand why we have this new added degree of freedom in the order of the axis, weights, etc ...

[ntessore]

Thanks @JaimeRZP! merging now