Akka.DistributedData.LightningDb: move durable folder creation outside of actor constructor in order to preserve stack trace / provide clearer error messages

[Aaronontheweb]

Affected libraries: Akka.DistributedData.LightningDb
Affected Versions: all

akka.net/src/contrib/cluster/Akka.DistributedData.LightningDB/LmdbDurableStore.cs

Lines 89 to 92 in abae81c

	if (!Directory.Exists(_dir))
	{
	Directory.CreateDirectory(_dir);
	}

In the event that this directory isn't writable by the process, the Akka.DistributedData.LightningDB.LmdbDurableStore will crash inside its own constructor and the real reason for the crash will be obscured by the error message:

Akka.Actor.OneForOneStrategy,Error while creating actor instance of type Akka.DistributedData.LightningDB.LmdbDurableStore with 1 args: (  keys : [shard-*]
  pruning-marker-time-to-live : "10 d"
  store-actor-class : "Akka.DistributedData.LightningDB.LmdbDurableStore, Akka.DistributedData.LightningDB"
  use-dispatcher : akka.cluster.distributed-data.durable.pinned-store
  pinned-store : {
    executor : thread-pool-executor
    type : PinnedDispatcher
  }
  lmdb : {
    dir : ddata
    map-size : "100 MiB"
    write-behind-interval : off
  }
)

We should really move this initialization to a messaging handler so we can clearly log a file permissions issue and make this easier for end-users to troubleshoot.