Assign np.nan only when the type is float

[delaossa]

This just prevents assigning np.nan to data fields in libE_output which are not float, but does nothing when the data is not float.
Possibly, we would want to assign default values to other parameter types such as int.

It addresses this issue #218