Problem with atomic replace when user does not own file

The atomicReplace function chowns the new file to match the ownership of the old file. This can fail when the file is owned by a different user than the FDB user, even if the FDB user has permissions to write both files.

For example, the FDB process is running as `user1`, which is a member of `group1`. And that the cluster file is owned by `user2:group1`. The cluster file is in `/var/fdb/fdb.cluster`, with the permissions `rw-rw-r--`, and `/var/fdb` has permissions `rwxrwxr--`. The process will successfully write to a temp file in `/var/fdb`, and the temp file will have the owner `user1:group1`. The process will then chown the temp file to `user2:group1`, which will fail.

We've run into this in a couple of different scenarios on Kubernetes, when changing volume permissions or running different containers in a pod with different users. I think we should find a way to make the atomic replace work in this scenario.

I believe this behavior was introduced in #1316, to address #1211.