repair 'make test-full': remove fails for coupled runs, cleanup start_bundle_coupled, remove outdated configs

[orichters]

Purpose of this PR

• the default is to use the current magpie folder also for oldruns. If that folder does not yet exist because no MAgPIE run was started, that is not a problem, but the script fails here. That shouldn't be, so simply create path_magpie/output.
• Remove message for people still trying to explicitly set --parallel mode. Now shows: Unknown flags: -p. Only available: -h/--help, -g/--gamscompile, -i/--interactive, -t/--test
• specify startgroup in coupled tests explicitly in case it was adjusted in start_bundle_coupled
• remove outdated config files as agreed with users by @LaviniaBaumstark. Fixes configs that fail start.R --gamscompile #1367
• fix stupid bug in xlsx_IIASA.R
• remove fallback to convGDX2MIF_REMIND2MAgPIE in reporting.R. If reporting fails, the run should as well
• As make test-coupled now actually uses reportingREMIND2MAgPIE since Allow appeding to cfg$output via scenario_config.csv; allow setting output also for coupled runs; update comments #1575, repairing make test-coupled requires repair convGDX2MIF_REMIND2MAgPIE pik-piam/remind2#550
• adjust the script that combines REMIND and MAgPIE data in reportingREMIND2MAgPIE.R to the code used in reporting.R to make tests more helpful
• add option to Makefile for running the full tests on slurm

Type of change

• Bug fix

Checklist:

• My code follows the coding etiquette
• I performed a self-review of my own code
• All automated model tests pass (FAIL 0 in the output of make test): [ FAIL 0 | WARN 0 | SKIP 6 | PASS 81 ]
• Most automated tests of make test-full pass. Output generation from test_06-output failed because of this memory issue, but the rest went through, so I think that is ok.