git/housekeeping: Improve metrics when repacking objects fails

When repacking objects fails during housekeeping we report a bunch of metrics relating to this as failed. We completely ignore the repack configuration though that led to the failure, which means that we would e.g. report both a "full" and an "incremental" repack as failed even though we actually performed a "cruft" repack. Furthermore, we do not take into account whether we actually tried to write bitmaps or the multi-pack index but also always report them as failed. This means that our metrics can be misleading at times, and that we cannot see whether a specific type of packing objects is more likely to fail than others.

Fix this error by bubbling up the repack configuration even when executing the repack command has failed. Like this, we can take the configuration into account even on a failed repack and only report these parts as failed that actually did fail.