Harvest discrepency log

[ghost]

During harvest of a collection, an isolated log should be created of failed SparQL requests, multiple most-specific types, missing required properties, duplicate literals with slight variation, and inconsistent language attribute use.
The log itself should probably be machine readable and contain as much relevant information as possible. The log may be emailed and or used in reattempts or other scheduled processes.

Size as assuming we won't be changing the existing SPARQL. Furthermore, this will occur only in the context of a full harvest and will re-create the log each time. Just put in hooks for logging whenever the various discrepancies are detected in the course of harvest.
Amending an existing log in place would be a future feature.