Dump metadata to recreate PTO after crash

[ghost]

If the PTO fatally crashes and takes with it the entire observation database, this is in theory no big deal, since all we need to do is re-run the normalisers and analysers. This may be true, yet the devil will be in the details. For example, observation set IDs will in general not be the same when an observation database is thus recreated. Jumps in observation set IDs can happen, for example, when observation sets are deleted ad then new ones added. Recreating observation sets just by re-running the analysers will then not recreate the same observation set IDs. This may invalidate URLs pointing to such observation sets.

I suggest therefore to develop a tools that can dump and restore the observation database from its metadata exactly as it is.