RSocketClient refactoring

[rstoyanchev]

RSocketClient is not yet in production so we've an opportunity to revisit things like the packaging, Javadoc, as well as where and how the client can be created.

One high level goal of this change is to draw a more clear contrast between RSocketClient and RSocket. There is a major update to the Javadoc of RSocketClient related to that as well precising the language to reflect the addition of LoadbalanceRSocketClient as a second alternative implementation since it was first created.

Secondly to emphasize them as as two distinct layers, in the proposed changes RSocketConnector is now left primarily for creating Mono<RSocket>, i.e. connecting to establish a live, transient connection. Separately RSocketClient.from(Mono<RSocket) allows wrapping a Mono<RSocket> source and uses it to establish (and re-establish) a shared connection for multiplexed use as needed.

The change also moves RSocketClient from the top-level package, where RSocket is for example, to the io.rsocket.core sub-package. For once RSocketClient is not on the same level as RSocket. It's only for sending, it works on a higher level than RSocket, etc. On a practical level this makes it possible to have static factory methods for the default implementation and for adapting a live RSocket.

[spencergibb]

Great clarifications. /cc @olegz

[OlegDokuka]

LGTM