fix(rust): Fix `sink_ipc_cloud` panicking with a Tokio runtime error

[Qqwy]

Fixes #13614

When writing ot ObjectStore-compatible storage in the IPC format, it seems like the .block_on calls inside the constructed CloudWriter might sometimes get called inside another block_on call. Tokio does not like this, resulting in a panic.

This PR resolves this issue by using block_on_potential_spawn in the necessary places instead.

[josevalim]

Hi @stinodego! I am sorry for the ping, but is there something blocking this PR or something we could do to get it merged? Thank you! :heart:

[Qqwy]

I'm still available to (help) fix(ing) the merge conflicts once more if needed, but it would be nice if it can be merged soon afterwards before new conflicts would arise 😅

[philss]

I think this is now conflicting with my fix from https://github.com/pola-rs/polars/pull/18027 I'm also available to help if needed :)

It seems that this fix would also fix the issue https://github.com/pola-rs/polars/issues/13163

[ritchie46]

I think this one can be closed as there has changed a lot since then.

[josevalim]

Is a fix still accepted? The functionality is currently broken and we would be happy users of it. Thank you :)

[ritchie46]

Yes, it is accepted. But I believe it was fixed by #18027

[philss]

I'm afraid this was not fixed yet. I could reproduce the error with the latest version from the main branch. Here is the backtrace: https://gist.github.com/philss/a81fab6f9cf85d23a0f1c364633e992b