Fix volume normalisation on Raspberry model

[ajj]

As noted in #815 the raspberry model needs to be doing volume calculation outside Iq function to get polydispersity working properly.

[pkienzle]

Should use the square() function for complex calculations:

     sfSS = psiS*psiS*sinc(q*(rL+deltaS*rS))*sinc(q*(rL+deltaS*rS));

becomes:

     sfSS = square(psiS*sinc(q*(rL+deltaS*rS)));

This makes life easier for the compiler since it doesn't have to figure out that the two subexpressions are identical, and saves four multiplies and a branch for each q.

Note that we now have sas_sinx_x as an available function instead of sinc.

[ajj]

This model doesn't have a single form factor volume as it actually is the sum of two form factors - the raspberry, and the free small particles.

Need to add polydispersity loop directly in this model as the auto polydispersity cannot work.

Longer term, should we implement models like this as composition models?