Investigate loss of precision in 2-way frequency observable partials

[DominicDirkx]

The code merged here:

#270

incorporates partials for the 2-way frequency observable partial (e.g. for PRIDE Doppler data). However, the resolution of the test of the partials seems to depend on the value used for the transmitted frequency, indicating a loss of numerical precision somewhere in the algorithm (of about 2 orders of magnitude when going from a fictitious value of 1 Hz transmission to 3 GHz)