Migrate to MonadUnliftIO

[k0ral]

• streaming-concurrency depends on monad-control to unlift functions to IO;
• streaming-bytestring indirectly depends on unliftio-core for the same purpose (via resourcet).

As a consequence, users of both streaming-* libraries have to mix MonadBaseControl IO and MonadUnliftIO in their code, which is quite messy. Not to mention that streaming-concurrency also indirectly depends on unliftio-core via streaming-with.

For the sake of consistency, would you accept a pull-request to ditch monad-control completely (respectively lifted-async), in favor of unliftio-core (respectively unliftio) ?

[ivan-m]

Resourcet is likely to be removed from streaming-bytestring (just as it has been with streaming itself).

How does streaming-with depend on unliftio-core?

[k0ral]

streaming-with depends on streaming-bytestring.
streaming-bytestring depends on resourcet.
resourcet depends on unliftio-core.

Of course, if streaming-bytestring drops the dependency on resourcet, then the issue vanishes.

[jkachmar]

I took a crack at this for fun and ended up with these changes.

Notably I'm less interested in it from a resourcet-compatbility standpoint and more from a "I'm concerned about the safety of monad-control" standpoint.

I'm trying something similar with streaming-with and unliftio as well just to see how ergonomic things end up staying.