Protect against accidental creation of duplicates

[rybesh]

Currently the following not-unreasonable workflow can result in accidentally creating duplicates in PeriodO:

1. I create a new local database, and sync a period collection from the canonical dataset.
2. I add some new periods to this collection in my local database and submit a patch.
3. The patch gets merged. Great!
4. Now I add some more new periods to the collection in my local database, thinking that it is "in sync" when actually it still has skolem IDs for the merged new periods.
5. I submit a new patch, not realizing that it is a superset of my earlier, already merged patch.

Possible solution: give backends a GUID, store it with submitted patches, and before allowing patch submission from a backend, query for merged patches that resulted in new URIs, and update the skolem IDs (this assumes we are keeping that mapping around when we deskolemize).