Ensure that engine parameter is specified in every open_dataset call

[pont-us]

From xarray issues such as pydata/xarray#5329, pydata/xarray#5295, and pydata/xarray#5291, it is clear that it's dangerous to use xarray's open_dataset or open_mfdataset without explicitly specifying the engine parameter. Quote from this comment by an xarray developer:

The logic for autodetecting the engine has always been extremely fragile and the result depends on what packages are installed on the system.

In principle, any engine plugin installed in the user's environment (including ones we don't know of) could claim the file and do anything with it (including throwing exceptions and hanging indefinitely). So in every open_dataset or open_mfdataset call where it's possible, we should explicitly specify the engine parameter. If we have any calls where this isn't possible (i.e. we don't know in advance what filetype and/or engine we're expecting), we should consider on a case-by-case basis how best to mitigate unexpected engine behaviour. But I suspect that there will be very few or no such cases.