runtime tests

[loriab]

• adds BSD-3C license file closes license #10
• adds import structure so tests found at builddir and from installdir
• installs miniver (like versioneer only lighter) so that versions compute automatically from annotated tags. so git tag -a v0.3.0 -m "v0.3.0" for the next version, then git push upstream v0.3.0 to tell GH
• I've just minted v0.2
• introduced a fn that'll clear psi globals between tests. only deployed to one test file.
• How are sf_ea_ip and PyFockCI relating? Is only in transition to the other? Is sf_ea_ip currently the only cog in the future pyfockci universe?
• the bloch test_1 is the only test failing for me with development psi4

[shannonhouck]

Hi, sorry for the late response! To answer your question, sf_ip_ea is just the old package name-- I ought to change it over to pyfockci at some point for clarity!

Relative to the old Psi4FockCI code, PyFockCI is a code that provides an alternate approach to RAS-SF-IP/EA problems. The main difference is that it uses explicit hard-coded diagrammatic equations instead of a string-based/Slater-Condon approach for the Davidson steps in the diagonalization. It's a lot less flexible than Psi4FockCI in terms of the types of calculations it allows the user to perform, but in theory the diagrammatic approach should provide some speedup for very large systems (especially since PyFockCI can use density fitted integrals!). Ideally, I'd like to maintain them both, since Psi4FockCI lets the user perform an arbitrary number of spin-flips or IP/EAs, and allows for a greater variety of excitation levels.

I'll go ahead and fix the Bloch test; thanks for mentioning that! My guess is it's probably a difference in localization of the ROHF solutions between Psi4 versions, and I just haven't gotten around to updating it.