Store atom indices in ICRMSD dataframes

[mattwthompson]

From YDS I'm getting artifacts that are mostly useful but lack complete context:

$ head icrmsd.csv
,Bond,Angle,Dihedral,Improper
36975451,0.018579918334065595,2.803874569684111,11.67325536340797,6.004583328125764
36975452,0.01464782329727232,2.4066300918019596,15.361599111771518,7.038132838295687
36975453,0.01745440114119386,2.8658365564849784,12.325534818089592,6.834211449415355
36975454,0.0175933534460023,2.87254936926404,11.688686125733243,6.55216279079066
36975455,0.014525356950860548,2.219845637528826,11.08703272346969,5.7706638927499
36975456,0.018346739909489835,2.924661503227351,11.796789273892339,5.75333320102709
36991787,0.013226071277157065,1.0230487656285963,1.724340888254379,0.2494337741367869
36991788,0.012125714100575441,1.6300154173887207,17.78669206620193,0.8209735447427816
36991789,0.014462730369253482,1.3627500841921842,1.3357387216048286,0.28537742436397434

I forget if these are summed over all {bonds|angles|propers|impropers} in each molecule. If it is, they need to be split out, if it is not they need to be tracked to particular tuples of atoms so that we can more helpfully identify problematic chemistries.

[mattwthompson]

I now don't think this is worthwhile. This would be maybe 50-200 key-val pairs for each molecules and balloon the dataframe size by default. I think if #108 can be done without this I will close this.