Not to be flippant, but the cheapest way is to not buy a toolbox just for the functions you need, but simply write your own functions.
Represent quaternion as a simple double variable Nx4, where N = number of quaternions in the variable. Same as MATLAB.
Then the quatmultiply, quatconj, and quatinv functions would be trivial to write. E.g., if we assume scalar-vector order and right-handed Hamilton convention as MATLAB toolboxes do, then assuming you have a later version of MATLAB that uses implicit expansion:
function qc = quatconj(q)
qc = [q(:,1),-q(:,2:4)];
function qi = quatinv(q)
qi = quatconj(q) ./ sum(q.*q,2);
function r = quatmultiply(q,p)
qs = q(:,1);
ps = p(:,1);
qv = q(:,2:4);
pv = p(:,2:4);
r = [qs.*ps - dot(qv,pv,2), qs.*pv + ps.*qv + cross(qv,pv,2)];
If you wanted to use the (evil) left-handed JPL quaternion convention (which does not match MATLAB), then you would change the + cross(etc) to - cross(etc) in the quatmultiply( ) function.
If you do decide to get a toolbox, keep in mind that the quaternion convention used by the Aerospace Toolbox is different from the quaternion convention used by the Robotics Toolbox. The Aerospace Toolbox seems to be geared towards coordinate system transformation quaternions, and the Robotics Toolbox seems to be geared towards vector rotations within the same frame.