Exposing enif_schedule_nif

[Qqwy]

As stated in issue #44's discussion, enif_schedule_nif is currently not exposed by Rustler because it is not known how to expose the function in a compile-time safe way,

as the NIF is supposed to return to the Erlang scheduler right after enif_schedule_nif was called.

My proposal would be to expose enif_schedule_nif in an unsafe context, to ensure that people that want to use it are mindful of if they are using it in the proper way.

In the future, it might be possible to wrap enif_schedule_nif in some way.
An example would be as a preemptive Iterator-consumer, to ensure that the NIF will be split in chunks of O(1), even when performing O(n) or O(n^2) etc iterations.

[hansihe]

I totally agree with this, if there is an API we can't expose safely, we should do it unsafely with warnings.

If you want to give this a go, I would be fully open to pull requests. I might give it a go next week after I am done with my talk at PolyConf.

[hansihe]

Closing this issue in favour of #107, it contains more discussion