Intake_esm feature: Support for Kerchunk (reference) on s3 file system

Related Issue:
Intake_esm catalog hosted on an s3 file system which points to kerchunk files also hosted on the same file system cannot be accessed properly by existing intake_esm scripts.

Solution:
A small change to line 53 onwards in intake_esm/source.py:
```
if data_format == 'reference':
if 's3://' in urlpath:
import s3fs
xarray_open_kwargs['backend_kwargs']['storage_options']['fo'] = s3fs.S3FileSystem(**kwargs).open(urlpath, **kwargs)
else:
xarray_open_kwargs['backend_kwargs']['storage_options']['fo'] = urlpath
```

Alternatives:
There might be other options to consider/features in intake_esm I'm not taking advantage of, if anyone has any insight let me know.
The above change relates to a later section:
`fsspec.get_mapper('reference://', fo=ref ...)`
Where for a kerchunk file on s3, ref needs to be the S3FileSystem object not the string path.

Start by reading intake_esm/source.py around line 53 and trace the later fsspec.get_mapper('reference://', fo=ref ...) call. Verify how reference files are opened for s3:// URLs versus local paths, then confirm that an intake_esm catalog and its kerchunk files on the same S3 filesystem can be accessed while existing local behavior still works.

Written by the indexing model from the issue text.