document `config.HF_DATASETS_OFFLINE` and precedence

https://github.com/huggingface/datasets/pull/1976 implemented `HF_DATASETS_OFFLINE`, but:
1. `config.HF_DATASETS_OFFLINE` is not documented
2. the precedence is not documented (env, config)

I'm thinking it probably should be similar to what it says https://huggingface.co/docs/datasets/loading_datasets.html#from-the-huggingface-hub about `datasets.config.IN_MEMORY_MAX_SIZE`:

Quote:
> The default in 🤗 Datasets is to memory-map the dataset on disk unless you set datasets.config.IN_MEMORY_MAX_SIZE different from 0 bytes (default). In that case, the dataset will be copied in-memory if its size is smaller than datasets.config.IN_MEMORY_MAX_SIZE bytes, and memory-mapped otherwise. This behavior can be enabled by setting either the configuration option datasets.config.IN_MEMORY_MAX_SIZE (higher precedence) or the environment variable HF_DATASETS_IN_MEMORY_MAX_SIZE (lower precedence) to nonzero.

Context: trying to use `config.HF_DATASETS_OFFLINE` here:
https://github.com/bigscience-workshop/Megatron-DeepSpeed/pull/48
but are uncertain if it's safe, since it's not documented as a public API.

Thank you!

@lhoestq, @albertvillanova