bugfix for accessing global environment during testing setup

When the R session is new, i.e. ogs5_keywordlist is not in the global R-studio environment, the testing leads to aborting the R-session. Somehow, I think this is linked to the call .GlobalEnv$ogs5_keywordlist <- ogs5_get_keywordlist() in line 26 of ogs5.R. I hope to find a more elegant (and bugfree) solution to this.