Generalize specification of default initial parameters for variational ansatzes

[kevinsung]

The currently implemented ansatzes have default initial parameters derived from adiabatic evolution starting from an eigenstate of the one-body term. However, we'll want to try out other things, such as starting from different initial states. I'm starting to think that VariationalAnsatz should not have a default_initial_parameters method, and that initial parameters should be generated by separate functions.