Speed up Vasprun parsing some more

[kavanase]

Me again.
From further profiling and playing around, I found I could speed up _parse_vasp_array (one of the main bottlenecks when using parse_dos = True (default), parse_eigen = True (default) and/or parse_projected_eigen = True (False by default)), using numpy's parse from string function.
e.g. parsing a SOC defect supercell vasprun via doped with these updates (with parse_projected_eigen=True to get eigenvalues/magnetisation) decreases parsing time from ~8.5s to ~4.8s.

All changes here should be covered by tests already in the codebase.

[shyuep]

Thanks. but I don't think we need to use string concat? I believe np.loadtxt would be able to handle the text without concat.

[shyuep]

Example:

In [1]: import numpy as np

In [3]: np.loadtxt(["1 2", "3 4"])
Out[3]:
array([[1., 2.],
       [3., 4.]])

[shyuep]

Even better, no reshaping needed.

[kavanase]

Ah yes! Good points. Done ⬆️