The bridge discovery participant receives lots of mDNS events.
Previously, all events that contained IP addresses of potential bridges
were actively contacted using HTTP requests. On some systems eventually
the long polling stops due to too many requests.
With this change, we
* only consider mDNS events where the name property starts with "Bosch
SHC"
* cache already discovered bridges so we don't have to contact them over
and over again
* make sure that this happens in a thread-safe manner because the mDNS
events are handled in individual concurrently running threads