[FIX] force spawning pool for MacOS

[burtenshaw]

This PR gets multiprocessing to work on mac os:

from datasets import load_dataset

ds = load_dataset("fka/awesome-chatgpt-prompts", split="train").take(100)

ds = ds.map(lambda x: x, num_proc=4)

ds.push_to_hub("burtenshaw/dataset-test", num_proc=4)

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[lhoestq]

we use multiprocess and not multiprocessing in order to support a wider variety of functions to run in parallel

moreover I'm not sure this is the right fix yet, I think this needs more investigation

[lhoestq]

After more discussions on slack, we can switch the default to spawn.

Let's use multiprocess instead of multiprocessing and maybe add a check to apply this only on Macos ?