Use delegating repository to get a session from for hst model loading and not a jcr session from hstconfiguser session pool

It turns out, that very rarely, in clustered environments, a session keeps a wrong cached hierarchy of a node that has been moved. This is not fixed with a session refresh

Newly created sessions do not suffer from this. Hence, using delegating repository to get a new jcr session and log it out at the end is safer in clustered setups