fix MAGICC7 variables to reference run

[orichters]

Purpose of this PR

• somehow, MAGICC7 does not produce the exact same output as the reference run for all timesteps before cm_startyear, see less /p/projects/remind/modeltests/remind/output/SSP2-PkBudg650-AMT_2024-07-11_14.44.30/log.txt.
• Many of the deviations are very small, well below 1%, while some specific Aerosol Forcing variables in the 90th percentile are as big as 40%.
• part of https://github.com/remindmodel/development_issues/issues/220

Type of change

• Bug fix

Checklist:

• My code follows the coding etiquette
• I performed a self-review of my own code
• I explained my changes within the PR, particularly in hard-to-understand areas
• I checked that the in-code documentation is up-to-date
• I adjusted the reporting in remind2 where it was needed
• I adjusted forbiddenColumnNames in readCheckScenarioConfig.R in case the PR leads to deprecated switches
• All automated model tests pass (FAIL 0 in the output of make test): [ FAIL 0 | WARN 0 | SKIP 6 | PASS 83 ]
• The changelog CHANGELOG.md has been updated correctly

[LaviniaBaumstark]

why is this PR still open?

[gabriel-abrahao]

I still haven't managed to find out why this is happening, but for the time being, just fixing should be good enough.