first node is not <clustername>-1

and don't log some errors, useless (fills up log files quickly)

to avoid growing the vagrant shared folder size exponentially

- remove globals probes_sets, probes_names_maps
(to avoid shared data between processes)
- probes_names_maps now member of HDF5_Collector
- remove get_probes_list(), all client code uses get_probes_names()
- get_probes_names() now takes start_time, end_time, so that the list
of probes returned is valid according to the given timeslot
- get_probes_names() gets data from the hdf5 files
- rest server: factorization, bugfixes, and calling get_probes_names()
with start_time, end_time

+ get them from the app context instead of copying them in the request
context on each request

- use different zmw contexts
- init driver context in Process.run()
- move driver work to work() instead of run()
- add conf option for driver to driver_manager zmq socket name