Remove our own tokio runtime and use the one provided by pyo3-async-runtimes

[kylebarron]

cc @ion-elgreco ; any thoughts? You previously mentioned here that we should check for forked processes. I'm curious if you've considered using the runtime provided by pyo3-async-runtimes?

Closes #419

[ion-elgreco]

I believe the forked processes checking remains, pyo3-async also uses Tokio right?

Pyo3-async didn't make much sense for delta-rs because all methods in python are sync, and the rust code is async

[kylebarron]

I believe the forked processes checking remains, pyo3-async also uses Tokio right?

pyo3-async-runtimes does use tokio, but I don't think it has as nice of a check as was here. I figure tokio will just silently stop working if the python user tries to multiprocess?

[ion-elgreco]

I believe the forked processes checking remains, pyo3-async also uses Tokio right?

pyo3-async-runtimes does use tokio, but I don't think it has as nice of a check as was here. I figure tokio will just silently stop working if the python user tries to multiprocess?

It might deadlock or panic actually if you use fork

[kylebarron]

Indeed this might fail when using fork-based multiprocessing, but we can wait to figure out how to document that when the first user complains.

[oelhammouchi]

@kylebarron am I to understand that obstore is not fork-safe? I'm running into this issue right now.

[kylebarron]

I think that yes, obstore is not fork safe. Do you have a suggestion of where to document that?