np.float64('nan') is not handled correctly

[gdementen]

Excel prints a 65535 instead (as in #15).

This is because of the following lines in prepare_xl_data_element (

xlwings/xlwings/_xlwindows.py

Line 1040 in 171b107

def prepare_xl_data_element(x):

)

elif np and isinstance(x, np.generic):
        return float(x)

In that case, x is converted to a Python builtin float('nan'), but not to "".

I will try to produce a PR to fix this next week.