Add interfaces for PETSc/SLEPc routines

[Peter Hill]

The eigensolver uses lots of PETSc and SLEPc functions which, unfortunately, don’t have interfaces defined for some reason.

We should define them, and perhaps contribute upstream.