Potential confusing issue with `frequency_domain_source_model`

To run on a BNS (i.e. to use tidal parameters), one needs to specify

frequency-domain-source-model = lal_binary_neutron_star

I think (I'm currently testing) that if you run with a waveform IMRPhenomPv2_NRTidal and the default frequency-domain-source-model=lal_binary_black_hole, then it will just return the prior on the tidal parameters, which is obviously annoying.

Options

1. Error if the waveform-approximant contains tidal (in any case) and frequency-domain-source-model=lal_binary_neutron_star.
2. Print a warning in the above case
3. Automatically set the frequency-domain-source based on the waveform approximant

Thoughts? @colm.talbot @isobel.romero-shaw @shanika.galaudage @charlie.hoy

My personal preference is option 1, with improved logging and documentation about the frequency domain source model.