Add options for initial guess and refactor SCF convergence algorithms

[agoetz]

In general the DIIS based SCF convergence with SAD guess implemented in QUICK works well and is efficient.

In some rare instances of specific combinations of molecules and basis sets the SCF can converge to a solution that differs from other quantum chemistry codes. We have seen this for cytosine HF and B3LYP single point energies with the STO-3G basis set and the cc-pVDZ basis set (not other basis sets) that matched some versions but not other versions of other codes. At this point it is not clear whether QUICK and/or which versions of other codes converge to the correct or wrong solution.

In any event we should provide additional options for initial guess generation

• core Hamiltonian guess (not a good guess but nice to have as reference)
• superposition of atomic potentials (SAP)

and revisit the SCF algorithms

• refactor / clean up
• add second order SCF