Better starting states for Arnoldi method in julia_trotter

[abhijithjlanl]

The bond-energy and the energy error scripts in julia_trotter use the Arnoldi method to find ground states of Trotterized unitaries without building full dense matrices. This quality of this computation depends on the starting state and currently the HF starting state is used.

A better options would be either,

1. Find the ground state of the true Hamiltonian and use that as the starting state.
2. Repeat the computation for some random states including the HF states and report the minimum.