pull: Add --per-object-fsync

pull: Add --per-object-fsync

This is the opposite of
https://github.com/ostreedev/ostree/issues/1184

Motivated by OpenShift seeing etcd performance issues during
OS updates: https://github.com/openshift/machine-config-operator/issues/1897

Basically, if we switch to invoking `fsync()` as we go, it makes
ostree performance worse (in my tests, 31s to write 2G versus 7s if we
delay sync) but it avoids *huge* outliers in `fsync()` time for etcd.