Need to enable correct post-1.4.11 PersistenceIds Akka.Persistence.Query support

[Aaronontheweb]

set AllocatesAllPersistenceIDsPublisher => false; so TCK specs pass

we're going to need to, in the future, change this plugin so these specs can pass.

Originally posted by @Aaronontheweb in #129 (comment)

[Aaronontheweb]

The issue is that currently, the PersistenceIds queries in the Akka.Persistence.Azure Journal are implemented using what was originally included inside the SQLite journal, which is bad an doesn't work across multiple nodes:

Akka.Persistence.Azure/src/Akka.Persistence.Azure/Journal/AzureTableStorageJournal.cs

Lines 48 to 49 in 51254a9

	private readonly HashSet<string> _allPersistenceIds = new HashSet<string>();
	private readonly HashSet<IActorRef> _allPersistenceIdSubscribers = new HashSet<IActorRef>();

Right now this works by having the journal track all persitentIds in-memory based on what ids are observed by the journal. Historically, we know that this is extremely memory-inefficient and fragile as it doesn't work for persistenceIds that aren't seen by the current node. This needs to be rewritten to query all ids, authoritatively, from the database instead.

[Arkatufus]

This is not possible to do on Azure because we did not store any ordering "column" in Azure, which is needed to implement the "AllPersistenceId" and "AllTaggedEvents" query

[talismax]

If I use this persistence plugin in a cluster sharding scenario with Persistence.Query, i.e.

`public void StartListeningForPersistenceIds()
{
    var materializer = Context.System.Materializer();
    var journal = PersistenceQuery
        .Get(Context.System)
        .ReadJournalFor<AzureTableStorageReadJournal>(AzureTableStorageReadJournal.Identifier);

    journal.PersistenceIds().RunForeach(ProcessPersistenceId, materializer);
}`

... will I encounter this bug? If so, any amazing workarounds?