`ScanSink` is not threadsafe

[isaac-weisberg]

There if of course no direct guarantee that operators should be thread safe in the Rx realm, and it is a matter of implementation, but at the same time the precedent of having locks inside Sinks has been set a long time ago - just look at the date on SynchronizedOnType, and it's of course used in a lot of places.

ScanSink maybe probably definitely should be internally synchronized since it owns a mutable state...

[freak4pc]

Hey @isaac-weisberg, thanks for raising this!
I think it's easier to reason about this with a real use-case and not theoretical thread-safety,
Do you have a use case / thought of where this poses an issue, or a code example where this would cause a problem?

I'm not trying to be push-backy, just trying to be careful with changes (especially locks) given how code-stable and "mature" Rx is, and how wide-spread its usage is in many apps that probably dont necessarily face issues with this.

Thank you 🙏