Provide parallel queries

[leoluecken]

If a single instance is not responsive or the IP quota is exhausted, this blocks batch queries, if these cannot be launched in parallel. One way to implement this would provide a parallel version of each method, that can take a list of hosts and corresponding arguments (plus a timeout) and returns a result object, which may be checked for completion.

Internally, it would be good to use the existing clients and only launch one or a limited number of requests per host. Perhaps, by having one worker per host?