coreg.BiasCorr warns for all-nans but doesn't fail

[erikmannerfelt]

BiasCorr should arguably fail if it encounters a DEM with only nans. Right now, a numpy warning occurs which is not very useful in identifying the problem.

Minimal example:

In [1]: import xdem

In [2]: import numpy as np

In [3]: biascorr = xdem.coreg.BiasCorr()

In [4]: dem1 = np.ones((50, 50))

In [5]: dem2 = np.ones((50, 50)) + np.nan

In [6]: biascorr.fit(dem1, dem2, transform=None)
/home/erik/.local/share/conda/xdem/lib/python3.8/site-packages/numpy/lib/function_base.py:380: RuntimeWarning: Mean of empty slice.
  avg = a.mean(axis)
/home/erik/.local/share/conda/xdem/lib/python3.8/site-packages/numpy/core/_methods.py:188: RuntimeWarning: invalid value encountered in double_scalars
  ret = ret.dtype.type(ret / rcount)

A sanity check should be added to make sure that the function fails instead of giving quite useless warnings!