Swap cdf_to_xarray to accept CDF instances, rather than filenames

[jphastings]

This change allows for CDF instances (eg. loaded from S3 buckets) to be converted to xarray Datasets, without being forced to retrieve the remote dataset, then load it from disk.

# Load the CDF from an S3 bucket or a URL in chunks
cdf = cdflib.CDF(some_cdf_url, s3_read_method=3)

# Use the loaded CDF when converting to an xarray
dataset = cdf_to_xarray(cdf)

fixes #312

I've implemented this as a breaking change to keep the function simple, but I'd be happy to rework that first argument as filename_path_or_cdf: str | Path | CDF and convert internally if you wanted to maintain API compatibility.

🚨 Breaking change 🚨

cdflib.xarray.cdf_to_xarray now requires a CDF object, not a filename.

- cdf_to_xarray("some.cdf")
+ cdf_to_xarray(CDF("some.cdf"))

[bryan-harter]

Ah yeah, this is a great idea! But that said, I think we should keep the function backwards compatible for now, since cdf_to_xarray is used in quite a few places. Maybe we could just add a check at the beginning of the function like "if not isinstance(file, CDF) then file=CDF(file)"?

[jphastings]

I’ll make the change and push shortly 😊