C#: Use * ID for Locations in buildless.

[michaelnebel]

In this PR we change the C# extraction to use * IDs for Locations. This is a recommend pre-requisite for starting the work on C# overlay support.

It turns out that there is a complication in relation to the "empty" location: There is a build in assumption in the QL library that we can refer to a unique single location for representing an "empty" location. That is,

• The value has to be present as much of the QL library assumes that the "empty" location can be used in case there is no other "real" relevant location available.
• To avoid severe performance issues (combinatorial explosion) the empty location needs to be unique.

Until now the empty location piggy backed on the fact that

• At least one generated location (with empty filename and 0 as positions) would be created on extraction.
• The trap importer would ensure de-duplication of these types of generated locations (ultimately resulting in a single such location).

As a part of this PR we now extract a single synthetic "empty" location to be used to represent the empty location in the QL library.

DCA

• It appears that the analysis time has increased slightly and especially for the mono repo. However, it is worth noticing that the end-to-times are approximately the same. An example: For mono the analysis time increases by 20% (which corresponds to approximately 20s). However, the trap import time is reduced around 20s as well - I suspect this is because without * IDs the trap importer needs to do a lot of location de-duplication (for mono there is approximately 17m source location tuples, where 5m are duplicates - which are removed due to labelling on main). Correspondingly the increase in analysis time can be explained by the more bloated location table.