U/jrbogart/allowing duplicate relpath

[JoanneBogart]

Don't require relative_path to be unique in case the dataset to be registered has old_location == None because there is no danger of overwriting.
Addresses the second part of issue #156

[stuartmcalpine]

Looks ok ... just a check...

• Could raise a warming rather than printing (see line 825 for example
• Should the archived case still behave as a warning. I thought for archived datasets, no matter what the relative path is reserved for the dataset possibly coming back

[JoanneBogart]

I've addressed your first point (thanks for the reminder about using warnings).

For the archived case, there is still no danger of anything being overwritten because the call to register isn't doing any writing. There is another problem, though, which is that, if the other dataset was archived, as a practical matter the new one is also. I could update the status to say "archived". But I think there will be other issues along these lines with datasets coming from GCRCatalogs, e.g., if one dataset is based on another and the underlying one is archived, the other one is, too, effectively. I'm going to leave this alone for now; when we implement archiving we'll need to revisit these situations.