Use I(q,t) for fitting instead of S(q,w)

**Is your feature request related to a problem? Please describe.**
We have been asked to enable the use of I(q,t) as the observable in MDMC refinement as an alternative to S(q,w).

**Describe the solution you'd like**
Since we already calculate the intermediate scattering function, it should be possible to add an observable name or a keyword argument which will allow us to switch from S(q,w) to I(q,t) without adding new calculations to the code.

**Describe alternatives you've considered**
We could insist on calculating S(q,w) from the input I(q,t) instead, but the cleanest approach would be not to change the reference input data, but to match the output of the code to whatever input we have been given.

**Additional context**
This can be combined with #1685