Allow existing coordinates on "join_new"

**Is your feature request related to a problem? Please describe.**
Ok so this one is a bit more complex, and may be too specific to be implemented here. I'll let you decide. In my database, there are some files where "member_id" is a dimension and some others where it's not. I want to have "member_id" in the catalog for searches and I want to be able to aggregate the output so that `member_id` is a dimension in the datasets of `to_dataset_dict`.

For this I use a "join_new" aggregation, which cover the the second case. However, when "member_id" already exist as a dimension in the file, this fails. For these cases, I should have used a "join_existing" aggregation.

**Describe the solution you'd like**
I'd like that when using "join_new" and the coordinate already exists on the dataset, no error is raised.

**Describe alternatives you've considered**
Two catalogs? Split all my files on disk? Remove `member_id` from the aggregation control?

**Additional context**
I tried something on a branch. It works by modifying `source._expand_dims`. If the dimension exists and has the correct coordinate, no expansion is done. See my changes here:

https://github.com/Ouranosinc/intake-esm/blob/4aa37fda43139b21b4feecf7457dd85f36c12856/intake_esm/source.py#L96-L117

and also further down, lines 232-239, so that iterable fields are passed correctly.

Is this an acceptable feature?

Read intake_esm/source.py around lines 96-117 and 232-239, starting with _expand_dims and how iterable fields are passed. Compare the referenced branch changes with the existing join_new path, then verify that an existing dimension with the correct coordinate no longer raises an error while aggregation still exposes member_id in the dataset output.

Written by the indexing model from the issue text.