Better control over printing of CONECT records to PDB files.

[drroe]

Previously, CONECT records were printed for all bonds if conect was specified, or only HETATM residues if pdbres/pdbv3 had been specified. This PR adds a new keyword, conectmode, which allows more direct control over the printing of CONECT records:

conectmode <m>  : Write CONECT records for <m>='all' (all bonds), 'het' (HETATM only), 'none' (no CONECT).

Manual updated and test added.

[AmberJenkins]

The PGI build in Jenkins failed.

[AmberJenkins]

The PGI build in Jenkins failed.

[drroe]

@swails So the PGI failure seems like the old machine-specific one. I notice that the Jenkinsfile changed from PGI requiring batwoman to intel-cpu - can we change it back or is batwoman gone? It seems like the one that passed in fact ran on Batwoman, while the failed one seems to have run on Green Lantern.

Also, the Pytraj failure can probably be fixed by sourcing the cpptraj.sh script (to set CPPTRAJHOME so the J-coupling action can find the data file it needs) that is generated by cpptraj's install. Is there an easy way to do that? I looked at the Jenkins files, but I think all that logic is in the docker images right?

All that said, I've independently verified that everything builds and tests with PGI, so I think this is safe to merge.

[swails]

Batwoman is a very old machine (ca. 2011) with a Xeon processor. The other Intel CPU machines all have modern i5s (3 machines with 9400F). I'm a little wary of tying the PGI build to the oldest machine in the array since if it goes down, no build can finish.

When the machine-specific issue first arose, the two machines in the Jenkins array was an AMD FX-6100 (ca 2012) and a Xeon X5675 (ca 2011), and it only worked on the Xeon.

Give me a few days to dig a little and see what I can find. Alternatively, I can grant you ssh access to Green Lantern if you wanted to dig, as well.

[drroe]

@swails I think this warrants its own issue: see #841.

I've been able to reproduce the failure at CHPC Utah; helpfully they have both an architecture that PGI-compiled cpptraj passes for, as well as one it fails for. I'll continue debugging there for the moment.