Allow concurrent startup of Nuxeo instances

In cluster mode, the document repository and the directories are initialized non-concurrently in a cluster-wide critical section. When a cluster node attempts to initialize its repository (or a directory) and another node is already doing the same thing, it will wait for 1 min for the cluster-wide lock to be released and do its own initialization. If this timeout expires, then initialization fails with an exception. The following two nuxeo.conf properties can be used to change this timeout: org.nuxeo.repository.cluster.start.duration=1m org.nuxeo.directory.cluster.start.duration=1m In case where there's a startup crash while a lock is held, it may be necessary to manually cleanup the key/value store of its locks. The keys corresponding to the locks are visible when using Redis with KEYS nuxeo:cluster:* for instance nuxeo:cluster:start-repository-default or nuxeo:cluster:start-directories . For a MongoDB key/value store, the keys are stored in the collection  kv.cluster