C#/Java: Content based model generation improvements.

[michaelnebel]

In this PR we make the following content based model generator improvements stated below.
With these changes we are now ready to re-generate the models for the Java SDK using the "mixed" query mentioned below.

• Instead of completely excluding an API from content based model generation, if it overall has "too many" summaries, we now consider the number of summaries pr. parameter instead, and only exclude an API and parameter combination in case it could yield more than three summaries.
• The requirement for when we accept synthetic fields has been made less tight. Access paths containing a synthetic field should no longer be identical to form a chain of synthetics. The synthetic fields should now just be mentioned in the same order in some other access path (but a chain should still be formed).
• Content based summaries now have their own provenance dfc-generated.
• The lifting logic for the content based summaries has been improved.
• A new "mixed" model generator query has been introduced: If at least one content based model is generated for an API we use the content based model; Otherwise we attempt to generate non-content based models and use them instead. A query for mixed neutrals has also been introduced.
• The model generator script has been updated such that we can create mixed summaries/neutrals.

[michaelnebel]

DCA looks good!

[hvitved]

LGTM, some minor comments.