use Java CompletionStage instead of Scala Future in pekko-persistence Java APIs

[pjfanning]

• based on Don't use Scala Futures in Java APIs #1419
• breaking change so will target Pekko 2.0.0
• it might be useful to do some follow up changes but I think this is enough to start with

[He-Pin]

map(option)(org.apache.pekko.dispatch.ExecutionContexts.parasitic)

[mdedetrich]

@pjfanning Regarding parasitic, for future reference whenever you do a map operation on Scala's Future, always pass in the explicit ExecutionContexts.parasitic. If you don't do this, the block inside of the map has to go into the implicit ExecutionContext (usually some kind of fork join pool), and that context switch is unnecessary overhead for what typically happens inside of a map.

Context switch is fine (and needed) for long running async operations, but those are typically composed with flatMap (since they would also be running in some kind of Future). When it comes to long running blocking that aren't inside of a Future, you would use the blocking/specific thread based dispatcher instead

[pjfanning]

@He-Pin I made a couple of changes after your review was made including fixing an issue. Do you still think it is ok to merge as is?

[He-Pin]

Lgtm