* 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 the reference scenario (for calculating the BECCS bound, see previous commit) from input_ref.gdx to the values of the own previous nash iteration.

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

- added documentation to the output script.
- updated the multi-comparison-matrix.
- updated the prefix used for coupled runs.
- correctly handle the number of cores to be used in parallel.

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.

Script now returns after saving RDS files when in reporting mode.

- remove *explicit* flags for vintages and learning, essentially
switching those to TRUE by default.
- save and load RDS files to a subfolder in the output directory to
not clutter the output dir too much.
- add and group all RDS file save that are related to reporting.

This reverts commit 9b510e4b.

edgeTrpLib, see b6f2a2c6.

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.