cluster: broadcast message to many actors of same kind about protoactor-go

[sergeybykov]

We discussed similar ideas at length in dotnet/orleans#610. I tried to spell it out in dotnet/orleans#610 (comment). As was correctly pointed by @gabikliot, cases 1 and 2 semantically are pretty much equivalent to streams. Except we don't efficiently broadcast stream messages today.

So far we only implemented case 3 dotnet/orleans#2444, which is not really about broadcasting, but was mentioned in that discussion.

[stuartcarnie]

We could also expand the TellMany API to include an additional option controlling the delivery behavior if actor is not started:

• ActorAutoStart
• ActorSkip

[rogeralsing]

cc @ReubenBond, @sergeybykov.
I assume you two have explored this area, batching/grouping broadcast messages for virtual actors across partitions and location nodes?

Thoughts? valuable? problems?

To clarify the Go syntax. Names are the same as your Grain ID's. "Kind" is the actor kind. Go has no generics, so well, we have strings to notify what kind of Grain we want to talk to.

[rogeralsing]

Thanks @sergeybykov for the feedback, much appreciated!