Use udev to generate content identifiers

Change the way we obtain content identifiers, so that there are bigger chance that these are based on partition/subvolume UUIDs, instead of flakier device names.

As access to udev is tricky from the extractor due to sandboxes, the way we generate these content identifiers also had to change, so it is the indexer which creates a shallow nie:InformationElement in advance, that the extractors will take care of picking the content identifier from there, and further populating this content-side object.

This drops usage of libblkid, which does rely on cache files in order to make the information available without admin access, but these cache files have no guarantees of being generated in advance.