Minor update to make FermiDos more robust

[kavanase]

This is a minor fix to the get_doping method in FermiDos to make it a little more robust, so that it can handle rare cases where there is only either 0 or 1 energy increments between the VBM and CBM indices (e.g. Dirac cone systems or toy test cases). This is done by starting the CB integral (for electron concentrations) from either the mid-gap index or, if mid-gap index is the VBM, from VBM+1 (= CBM index), and vice versa for the VB integral for hole concentrations.

[shyuep]

Thanks. Can you add a unittest for the functionality pls?

[kavanase]

Done! Didn't add one at first as it should be a relatively rare case where this actually matters (e.g. Dirac cone systems or toy test cases), but added now!
Some tests seem to be failing for unrelated reasons

[mkhorton]

Are you sure they're unrelated? I see e.g.

    def test_get_gap(self):
>       assert self.dos.get_gap() == approx(2.0589, abs=1e-4)
E       assert np.float64(2.3163) == 2.0589 � 1.0e-04
E         
E         comparison failed
E         Obtained: 2.3163
E         Expected: 2.0589 � 1.0e-04

Thanks for the PR!

[kavanase]

@mkhorton ah yes sorry! I just looked at one failing test and it seemed to be something different.

I've fixed that one now, and it's a nice validation as it now gives a Dos.get_gap() value which better matches the eigenvalue_band_properties result (which is typically the most reliable).