Fix/fit_single_crystal_parameters demo

[Gbowker]

See issue #445

This demo submits and runs fine in the develop build, however, when looking at the post-processed result, the model cannot converge on the right single crystal parameters, as shown in the plot:

First things first, the necking region is included in the data - surely this is affecting the convergence? In commit 3dae49b I've removed this necking region from the data.

@aplowman has already also fixed one part - the jacobian in formable:

-           left = jac_prod + damping * np.diag(jac_prod)
+           left = jac_prod + damping * np.diag(np.diag(jac_prod))

This makes the simulation actually start to converge:

However there is another problem - sensible parameter bounds have to be put into the code so that when searching the parameter space, parameters can't go negative as appears to be happening for the CRSS:

• Surely we can set bounds on the entire parameter space?
• @samuel-hg-engel says it might hit a "wall" where its in a local minima and wants to decend further to the global minima but can't becuase it has to go -ve to do so.