Modify https://github.com/jupyter-server/jupyter_server/pull/1367

[fcollonval]

This modifies jupyter-server#1367 in order to:

• allow only a valid hash algorithm from hashlib available algorithms
• don't require a new attribute on ContentsManager as contents manager are expected to update to the latest API supporting hash
• Make Mixin classes backward compatible as they may be reused downstream
• Don't change the purpose of _read_file (it only read files)
• Limit the number of new methods (only adds _get_hash)
• Don't set hash_algorithm except if hash is set
• Don't set the content if we only want the hash (it will avoid encoding the bytes for example)
• Don't read twice the content for notebooks

@Wh1isper feel free to cherry-pick my commits to your branch. I would strongly suggest removing the downstream test against jupytext because we have already very limited maintenance capability.

[Wh1isper]

This is a good fallback method. I had thought about using inspect.signature and I just want to get your opinion.

[fcollonval]

I thought about doing that once. But as the handler are instantiated for each request, I think the try/except is faster than signature inspection.