ENH: Make float dtype conversion consistent

[felixschmitz]

    out["hh_vehicles_value_a"] = fix_missings_make_float(raw_data["v010ha"])

Having a function for smallest int is fine; float should always be the default 64bit.

Originally posted by @hmgaudecker in #47 (comment)