Feature/remove opm

[einar90]

This PR removes the opm-common dependency from FieldOpt, along with the feature that relied on it: simulator deck parsing.

This was done for two reasons:

1. The DeckParsing feature is unreliable: it will often give strange/erroneous results. This stems from the way opm-common parses the deck (it is likely due to some part of the process that I don't fully understand, and may not actually be wrong if interpreted correctly.)
2. The opm-common library causes huge issues when attempting to compile on clusters.

Note: I tried to compile this on Maur, but all GCC modules < 6.2.0 no longer work (we've previously used 4.9.3). Using more recent versions compiles all dependencies just fine, but FieldOpt cannot be linked because the boost 1.59 module does not work properly with GCC modules other than 4.9.3.

I believe this branch should now compile just fine with the dependencies listed in COMPILING.md.

[Bellout]

Good work. Looks good, not many changes needed to extract the whole module, essentially only the conversion from blocks to spline. Was there not a functionality having to do with schedule-reading, control-setup and/or schedule writing that opm-common was supporting...? Anyhow, if it works, that is fine.

I did not quite understand the Maur comment: are your confirming it compiles on Maur, given som specific combination of (gcc, boost) libs?

[einar90]

We had that functionality, yes. It was in ScheduleParser, which translated the objects created by OPM when parsing the schedule into the structs used by Settings::Model. I.e. it was there, but it's removed now. It was very isolated, so removing it didn't require much.

About the Maur comment: The GCC 4.9.3 module, which we previously used to compile FieldOpt on Maur, is broken. The Boost 1.59 module and the openmpi 2.0.2 modules which FieldOpt also needs depends on the GCC 4.9.3 module to work, so we cannot use any other GCC module to compile.
However, the dependencies compile just fine using more recent GCC versions.
To conclude: I'm fairly certain that FieldOpt will now compile on Maur if the GCC 4.9.3 module is fixed.