Refdata throws exceptions in log if perlmutter is up and LRC is down

[MrCreosote]

If perlmutter is available and LRC is not, on a refdata stage request via the API:

• A refdata doc is created in mongo without an LRC section
• The perlmutter refdata staging job is started
• When it completes, the perlmutter job flow calls on_refdata_complete
• This is hooked up to call the lrc job flows's nersc_refdata_complete method in app_state.py
• the LRC job flow loads the refdata doc and tries to get the LRC section, which throws an exception as expected
• The exception is caught and the LRC job flow tries to save it to mongo
• The mongo refdata update method with throw an error since it tries to query by the refdata ID and the cluster id, but the document doesn't have the cluster id in it
• These errors will wind up in the server logs but the refdata document should be unchanged

I'm not actually sure if this is an issue or not. Maybe it's ok as is. It seems unclean though