Improve documentation and error catching for SimInspiralFD

Description of problem

When calling SimInspiralFD with a maximum frequency that is not the Nyquist frequency of a corresponding time-domain waveform, the deltaT passed to the TD waveform generator is inconsistent with the deltaT that is calculated after increasing f_max to the nearest power of 2. The end result is that the FFT'd data is produced with a deltaF that does not agree with the deltaF passed to SimInspiralFD as the FFT calculates freq->deltaF = 1.0 / ( time->deltaT * time->data->length ). When calling SimInspiralFD with an FD native waveform, these issues are not encountered.

This is the issue that was flagged up by @charlie.hoy in the PE call, see here. Neglecting this effect was shown to result in biases in PE.

When a user calls SimInspiralFD with a maximum frequency that is not Nyquist, we need to either trigger a visible warning or throw an error. The documentation must also be improved to clearly highlight this undesired behaviour.

Tagging: @patricia-schmidt, @colm.talbot, @aaron.zimmerman, @maria.haney, @vijay.varma, @jolien-creighton