Batch merge failure results in immediately abandoning pipeline and rebasing

When batch MR fails, Marge will emit output like It said Batch MR !22940 failed: CI failed! See pipeline https://gitlab.com/freedesktop-sdk/freedesktop-sdk/-/pipelines/1540642605. I will retry later...

However, it does not in fact retry later but instead immediately rebases one of the MR's which can result in dozen hours of wasted CI time. It should take a break at this point before doing anything further allowing for Gitlab retries or human intervention to kick in. Even if situation still after that is that Marge still cannot merge what was needed, it should not rebase but drop one of the MR's and try a new batch if there were more than two MR's in original batch.