Veronika Vodeb / ctools

The energy dispersion was so far neglected in ctmodel::get_obs.

Updated the model trimming so that the models that are outside an observation are offloaded to another model container and then re-appended after the evaluation.
Parallelization scheduling set to "static" with a chunk size of 1. This will slightly help with the issue in which multiple observations in a row do not contribute to the model cube and thus dividing the observations into "n-observations/n-cores" is not ideal.

Fixed the variable name in the header file to match the name in the ctmodel.cpp file.

Removed a small memory leak in trim_models that was introduced in previous commit.

Instead of making a full copy of all models and removing unwanted models, models are added if they overlap with the ROI in 'trim_models()'. This avoids unnecessary memory allocation/deallocation.

Implemented a change to remove all models that dont overlap the current observation when filling the cube. This speeds up the computation when filling from a large number of models.

Changed the on time for a given bin to be set from the observation being filled into the cube rather than the cube itself.
Reformat code (#2147)

Optimized computation intensive parts of ctmodel to reduce computation time.
Implemented openmp parallelization of ctmodel to reduce computation time.
Updated information logging (#2147)

Updated writing information to log file in order to prevent messy log file messages. This doesnt appear to impact the computation time.
Integrate code without modification of GammaLib (#2147)
Implement efficient model computation (#2147)
Update change log (#2147)

So far any exception occuring during the construction of the ctools were not catched. This is corrected now.

All likelihood ctools now derive from the ctlikelihood base class. This avoids code duplication and provides all likelihood tools with a uniform interface.