Feature/strain product lambda rework

This makes the StVenantKirchhoff and *ViscosityAssembler use lambdas to compute the stress from strain. I'd eventually like to make the NormalStressBoundaryAssembler accept a lambda, too (e.g. to allow for heterogeneous viscoelasticity rather than homogeneous elasticity).

Maybe the VariableCoefficientViscosityAssembler can disappear again completely now. Pretty sure I'm the only who uses it anyway.

Thoughts? Opinions? I don't currently consider this a great improvement so I'm open for anything, including outright rejection.