Add 2D Gaussian fitter to fit module

[ShinjiFujita]

Closes #109.

[astropenguin]

Thank you for splitting the updates. It seems that this PR is still in progress, but I would like to put our decision according to the previous DESHIMA meetings anyway. The goal of the issue #109 is to create a function decode.fit.cube that takes a cube DataArray (dimensions of (lon, lat, chan)) as an input, and return a DataArray of 2D Gaussian models as an output (with the same shape and dimensions as the input). We also discussed that we could use DataArray.curvefit so that a simple 2D Gaussian function can be applied to each channel simultaneously. So the possible design of the function would be:

def cube(
    cube: xr.DataArray,
    /,
    *,
    init_amp: float = 1.0,
    init_x0: float = 0.0,
    init_y0: float = 0.0,
    ...
) -> xr.DataArray:
    """Apply 2D Gaussian fit to each channel of a 3D spectral cube."""
    return cube.curvefit(
        coords=("lon", "lat"),
        func=gaussian_2d,
        p0=(init_amp, init_x0, init_y0, ...),
    )

I am not fully sure whether we could implement like this, but it would be appreciated if you could consider such function-oriented implementation for readability and reproducibility.