Demonstrate how to handle a metadata change without having to reharvest the complete index.

Created by: adamreichold

This allows the indexer and server to read old datasets after a metadata schema change so that the system continues to work until the next harvest replaces the datasets by their new versions. (This means we only have to keep this code around until the next complete harvest or at least we only ever need to support the last version instead of all previous versions if we limit ourselves to one schema change per harvesting period.)