Added replication for message store

[hawesie]

As discussed in Lincoln, I've added the ability for the message store to write to and query multiple databases. This supports the (external) deletion of content from the "main" message store database, but allowing queries to succeed on the replicated content. A little more information at https://github.com/hawesie/ros_datacentre/tree/hydro-devel/ros_datacentre#replication

[cburbridge]

I'd be happy to merge this since if we don't use this approach for replication, as far as I see the code changes wont stop the datacentre working as it does now. @marc-hanheide I agree that it adds dependency on network communication, which might be bad. Even if we do the writes to off-robot databases in separate threads it might get uggly.

[hawesie]

This now disables first attempt at replication and instead uses mongodump and mongorestore to do bulk replication. The aim is to use this to schedule replication when the robot is charging.

[Jailander]

there might be a conflict with this pull request (because of me. I am sorry) I edited and merged the datacentre.launch to add machine tags this works I have tested it, but in this pull request there is a line being added to the launch file that may cause trouble.