henry bond length drift checks; speed up on rotation and translation

[SimonEnsemble]

• in henry_coefficient, we now check for bond length drift and reset the molecule from a fresh copy if a little drift is observed; fail if drift is so severe that results thus far cannot be trusted
• sped up rotate! and translate_to! with broadcast now that fractional coords are in a matrix
• added a few more tests for rotate and translate to make sure bond length drift is not observed. H2S is a more rigorous test since it is nonlinear compared to CO2.

[coveralls]

Pull Request Test Coverage Report for Build 164

• 13 of 13 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.01%) to 87.397%

---

Totals
Change from base Build 153:	0.01%
Covered Lines:	742
Relevant Lines:	849

---

💛 - Coveralls

[Surluson]

Tests run and most everything looks good.
The docstring for bond_length_drift shows the default value for atol=1e-15 while the actual function has atol::Float64=1e-14.
And I do see some difference between atol's in the test, what's the reasoning behind that?

[SimonEnsemble]

@Surluson thanks, fixed doc string inconsistency. we just can't expect the atol to be so low after rotating. 1e-14 is still small enough. eventually we should use quaternions for rotations as I hear they are more numerically stable!