WIP basic intensity calculations

[ajjackson]

Start with the simplest possible incoherent-approximation calculation: fundamental transitions in the fully-isotropic approximation.

• Use Euphonic DebyeWaller data structure
• We need to calculate this ourselves to recover the intermediate mode-by-mode tensors.
• At this point we are not thinking too hard about q-points handling: require the user to pass an array of nominal Q2 values. (This could become a function over modes data.)

[ajjackson]

I've run into an interesting quantization error, resulting from a Mantid-Abins performance optimization.

Rather than calculate each fundamental mode intensity as Q² U² before binning to energy, we calculate the U² part for each mode, bin those, then apply Q² scaling to the binned spectra. (For higher-order modes this is Q^2n.) The terms can be separated linearly like this because of the isotropic approximation.

However, this means the Q used for intensity calculation can differ from the exact E-Q line of the component modes; it is the intersection of the E-Q constraint with the midpoint of the energy bin.

This is pretty tolerable in practice and can be reduced by using fine bins (which Mantid-Abins automatically does when orders 3-10 are enabled.) But it does lead to a small discrepancy between simple reference calculations and data dumped from Mantid-Abins; accordingly we will need a generous rtol in some of the test cases.