Can we clarify decode_cf option of open_dataset? #3020

Dea all,

I encountered a small unforeseen bug using the decode_cf of open_dataset. By only reading the doc string, I was not aware of the complete meaning of this parameter. Especially, that it overwrites other options.

Example:
ds = xarray.open_dataset(file_name, mask_and_scale=True, decode_cf=False)

The result is an unmasked and not sclaed dataset. decode_cf simply overwrites other options. See the code: https://github.com/pydata/xarray/blob/stable/xarray/backends/api.py#L308-L312

A simple solution would be to explain, that decode_cf sets mask_and_scale, decode_times, concat_characters, and decode_coords to False. But probably it would be more convenient to detect option conflicts like in my example and raise a ValueError.

What do you think? Do you prefer any of these options? I could write the little PR.