Make float interface run fully in float mode

Currently the prolate spheroid code is `double` only, and I believe I run some calculations in double mode regardless of input precision. If selected precision is low enough, the calculation should also probably just automatically select float mode, regardless of input data precision.