Split user provided and auto-generated metadata

ElasticSearch returns search results as JSON
files which have three top level entries
for us to parse:
"project", "file", and "metadata".
Change the class structure to match this
hierarchy. This means, that the mapping
from JSON to Java objects just works
and we can pass the data on to the
front-end in the same structure as
the back-end received it.

This allows the separation of the
"standardized" metadata which is provided
by the platform's users
and our implementation-detail metadata,
such as GitLab related data.
This way, no conflicts arise when a new
key is introduced to the metadata standard.