extend core multi-shell model to support spline functions.

[pkienzle]

Add a polynomial order parameter to the core multi-shell model to support spline functions

    sld_core
    radius
    sld_solvent
    form=['shells','ramp','quadratic spline','cubic spline']
    n_shells
    thickness[n_shells]
    sld_surface[n_shells]

Form=0 is a series of shells, like the core multi-shell model, otherwise the SLDs match at the interfaces, with form=1,2,3 as linear, quadratic and cubic splines. Use a ​monotone cubic spline for easier control (available in bumps.mono).

Using a generic polynomial spline,

  rho_i(r) =  A_i r^p + B_i r^(p-1) + …

the integral from 0 to r_max is

  f = f_core - f_solvent + Sum [h(rho_i, r_(i+1)) - h(rho_i, r_i)]

with

   h(rho,r) = (4 pi)/q^2 Sum_(k=0)^p  d^k[-cos(q r)]/dr^k  d^k[r rho(r)]/dr^k  /  (q r)^k 

For cubic polynomials, this works out to

   h(r) = (4 pi) / q^6  [ 
     - cos(qr) (Ar^4 + Br^3 + Cr^2 + Dr) q^4
     + sin(qr) (4Ar^3 + 3Br^2 + 2Cr + D) q^3
     + cos(qr) (12Ar^2 + 6Br + 2C) q^2
     - sin(qr) (24Ar + 6B) q
     - cos(qr) (24A)
   ]

The terms can be rearranged using the difference in coefficients between layers Delta A, Delta B, … so we only need one evaluation of h for each shell.