Use stress-based variational forms based on 2nd Piola

[finsberg]

Instead of defining variational forms through the strain energy we now define them via the second piola stress. This means that a variational form on the form

$$ \int \psi (F) \mathrm{d}x = \int \psi_{\text{passive}}(F) + \psi_{\text{active}}(F) + \psi_{\text{comp}}(J) \mathrm{d}x $$

which has been solved by performing derivative to get the residual, i.e

R = ufl.derivative(psi*dx, u, v)

will now be computed using the second piola stress instead, i.e

$$ \int S(C) : \frac{1}{2}C \mathrm{d}x = \int S_{\text{passive}}(C) +S_{\text{active}}(C) +S_{\text{comp}}(C) : \frac{1}{2}C \mathrm{d}x $$

using

R = ufl.inner(S(C), 0.5 * var_C) * dx

with

var_C = ufl.grad(v).T * F + F.T * ufl.grad(v)

This is useful since $\psi_{\text{active}}$ can depend on variables (e.g T_a) which in electro-mechanical simulations can depend on u in which case a derivative will trigger a chain rule operation which is not what we want, see e.g ComputationalPhysiology/simcardems#253

To make this work we also rewrite all the strain energy functions ($\psi$) to depend on C rather than F so that we can use S = 2 * ufl.diff(psi, C).