Reintroduce writing of git information to log file (latest commit tag, changes relative to the latest commit).

Allow providing additional SLURM parameters via cfg$slurmConfig. Parameters that are later provided interactively via start.R overwrite the cfg values. Closing #139

* Enable prepare_and_run.R to restart fixed runs (https://github.com/remindmodel/remind/issues/112)
* Enable start.R to restart non-optimal runs

change version of gdx being used for NDC target calculation (now input_bau.gdx) so that correct NDC targets are also calculated for non-SSP2 pathways

Add .Rprofile to the list of files that are copied into the results folder *before* prepare_and_run is called.

gdx neutral: clean scenarios in scenario_config_coupled_SSPSDP.csv and compareScenarios.R, add start and end time (of the job not of GAMS) to runstatistics.

Introduce the option to send output scripts for single runs to the cluster (comparison scripts will follow later)

Subsequent runs: do not save the timestamp to cfg$results_folder in RData files that are stored on REMIND's main folder. These files might be used to restart subsequent runs by hand. In this case the timestamp would cause the script to stop because the results folder already exists.

Refactoring start scripts: fix bug in sbatch command, remove scenario name from name of log.txt so the rs script can easily find it.

Refactoring start scripts: adapt coupling scripts, call SLURM directly without the detour via cmd files

Refactoring start scripts: - delete start_functions.R because all it's functions have been replaced by new ones
- fix bugs and correct formatting in other files

Refactoring start scripts: - moved the definition of the slurm strings from  submit() in submit.R to choose_submit() in start.R
- renamed the function choose_submit() to choose_slurmConfig() and moved it to the new file choose_slurmConfig.R

Refactoring: -start scripts (part 5): renamed submit_run(cfg) into submit.R(cfg), deleted conopt files

Refactoring: - part 2: restructure start scripts (try to minimize workload for the login node on the cluster when starting a REMIND run by moving code into scripts that are included in the slurm job)