Allow saving data with latest numpy version: don't use removed np.float function

Created by: Famlam

@wfranssen @smeerten If you prefer np.float32 (which doesn't seem to be documented, but does still seem to work) or another float##, please let me know. This was the simplest solution.

According to https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations the float function is identical to np.float See #145