Error check on 250 formulae

[tobigithub]

Hi,
I calculated errors for ~ 250 formulas, it took 22h, there are couple of issues for C&H and CHO.
https://gist.github.com/3903459 (copy paste into EXCEL or CALC and use autofilter).
I will send the XLS to you also. Some formulae were not calculated, because i stopped
the run after 22h, so empty values need to be evaluated in the future.

The worst offender speed wise is C9O with 8097.22 secs in AMG20121014
10064 isomers in MOLGEN DEMO calculated in 0.02sec.
That is an factor of 404860.79 besides the isomer numbers not matching which can be fixed.
The speed behaves like C =number of C atoms, O number of O atoms
AMG time [sec] = C_C_C_C/(118.1 + C_C + O_O + 1.975_C_O - 21.63_O - 21.71*C)
That means anything beyond C9 can become very slow.

For the C&H formulae only C7H2 did not work, see issue before, but I did not do an INCHIKey comparison for other structures.

For CHO formulae, some worked, some not, no relation to degree of unsaturation.
But oxygen,nitrogen, and other elements need to be implemented as you mentioned. I think those issues can be fixed, no problem.

I used the VisualVm profiler to find the worst class offenders.
70% of all times is spent in the signature methods,
signature.DAG.updatelayer() 30%
signature.DAG.computeVertexInvariants() 15%
AbstractVertexSignature.java and so on....

Cheers
Tobias