Refine I(Q,t) data

[MBartkowiakSTFC]

Description of work
This PR uses the MDANSE S(Q,w) calculator, but takes the F(Q,t) as the calculated observable.

The example script pgme-openmm.py shows that the calculation is possible, but does not construct the correct simulation that would match the experiment results. It still uses argon as the model system, so the refinement is not going to converge.

Fixes

1. Add a keyword argument to pick a different dataset from the MDANSE output file.
2. Add an arbitrary postprocessing function to normalise the calculated data to the first value for each Q.
3. Add a data loader that can read multiple text files and combine them into a single 2D dataset.

To test
Run the pgme-openmm.py script in the examples directory.

Question
At the moment the code takes F(Q,t) and compares it to the experiment results. The experiment result is most likely I(Q,t). Should we be taking F(Q,t)^2 to get I(Q,t)?

[oerc0122]

Is there a reason for this to be in a main guard?

[MBartkowiakSTFC]

MDANSEObservable could be set to run in multiprocessing mode. We want to prevent the subprocesses from starting their own simulations.

[ChiCheng45]

Question At the moment the code takes F(Q,t) and compares it to the experiment results. The experiment result is most likely I(Q,t). Should we be taking F(Q,t)^2 to get I(Q,t)?

I had a look at https://pubs.aip.org/aip/jcp/article/133/23/234506/950138/Reduced-mobility-of-di-propylene-glycol and it says that they obtain $I(Q,t)$ from a fourier tranform of $I_{m}(Q,\omega) = S(Q,\omega) \otimes R(Q,\omega)$ so that $I(Q,t) = I_{m}(Q,\omega) / R(Q,t)$. So in this case, the expt data is not squared?

[MBartkowiakSTFC]

I had a look at https://pubs.aip.org/aip/jcp/article/133/23/234506/950138/Reduced-mobility-of-di-propylene-glycol and it says that they obtain I ( Q , t ) from a fourier tranform of I m ( Q , ω ) = S ( Q , ω ) ⊗ R ( Q , ω ) so that I ( Q , t ) = I m ( Q , ω ) / R ( Q , t ) . So in this case, the expt data is not squared?

Our own scientists also published a document https://doi.org/10.1016/j.physb.2019.02.041 where I(Q,t) is defined as intermediate scattering function, which can be Fourier transformed to give S(Q,w).

So, unlike in some other scattering techniques (e.g. diffraction), it seems that F(...) and I(...) are the same thing.