git2cosul wipes local cache on ssh connection failure

[Yekut]

The current behavior of git2consul when ssh fails is to wipe the local cache and reclone (root cause seems to be the same as #53). This behavior is a bit unexpected (the user pretty much has no control over when this would happen), could be time consuming, and may cause trouble if the correct state of the repo rely on some preprocessing done after the initial checkout and the launching of git2consul. Would it make sense to improve error handling during a pull, and ignore ssh connection errors?