Apply sensible bounds to yield function fitting

[aplowman]

From the literature, many of the yield function anisotropy (fitting) parameters are close to one. We should apply sensible bounds to the fits. Bounds should be defined in the respective yield function classes as class attributes, but should also be overridable.

We can use the fact that the complex yield functions are often generalisations of simpler isotropic yield functions (e.g. von Mises). What parameter values are required to reproduce the simpler functions? Use this to guide sensible bounds.