Building of TSs with reactant and product species geometries

[nateharms]

Conformational changes like inversion of chiral centers and rotation about double bonds rarely occur when reactants go to TSs (e.g. a reactant containing a cis double bond will probably have a TS with that corresponding double bond being cis). This is similar to the templating idea proposed in #33, but with a variation. I'm proposing that we do the following when building a TS geometry:

1. Read in QM optimized geometries, if they exist, for reactants and products.
2. Perform conformer analysis and QM optimization for reactants and products that have no optimized geometries and read them in.
3. Using optimized geometries for reactants and products, construct a TS geometry using a tighter bounds matrix for bound atoms (because we trust these values)
4. Perform a conformer analysis on rotatable torsions only because cis/trans bonds and chiral centers probably won't invert.

This would require us to do the following:

• [ ] Create a database of optimized geometries for autotst.reaction to read from
  • Possibly build the geometry with RDKit if a reactant or product geometry isn't available but post a warning.
• [ ] Revise the way the bounds matrix is edited in autotst.reaction
• [ ] Revise the conformer analysis in autotst.conformer.systematic to not include cit/trans bonds and chiral centers

[nateharms]

Idea popped up while working on my review so I thought I'd at least write it down