Incremental TD3: incrementalize analysis within functions

[stilscher]

Closes #351

This PR contains the refinement for the incremental analysis which allows to reuse previous results for the unchanged nodes in changed functions. Instead of differentiating only between changed, unchanged, removed and added functions, changed functions can now be compared more detailed based on their control flow graph to determine the unchangedNodes, primObsoleteNodes and primNewNodes in the function. The unchanged nodes are tuples of old and new nodes that match and that only have incoming edges from other matching nodes. The primary obsolete nodes are nodes from the old CFG for which no match could be found and that suffice to be destabilized in an incremental analysis run. That means, that all changed nodes can be reached from the primary obsolete nodes. The primary new nodes are the equivalent to the primary obsolete nodes in the new CFG.

The changes of this PR include

• configuration option for activating this refinement
• change detection on control flow graphs for different versions of a function
• extending the preparation for the incremental analysis in TD3 to handle the more detailed change info (it should also allow to activate refinement Incremental TD3: reluctant function destabilization #369 at the same time)
• adapting the updateIds function to differentiate between unchanged and changed nodes in changed functions when updating the ids
• extending the script for incremental test runs on repository to reset the incremental data when starting and also measure the total internal runtime

[jerhard]

Added some comments on minor things. I did not have a deep look into the actual CFG-comparison yet. Maybe you can add some short comments there?

[stilscher]

I added comments for the compare algorithm and renaming as you suggested in the comments. Also I added the separate compareCIL module which uses the compare functions either from CompareCFG or CompareAST based on the configuration.

Changes requested in review were made.

[jerhard]

Should this printout be there? Maybe use tracing instead.

[stilscher]

This only extends the existing printouts during the incremental preparation in TD3 to make them more comprehensible in case of reluctant destabilization. I guess one should change all of them such that they are only output in verbose mode or use some existing tracing functions...

[sim642]

I think for the time being it's better if they aren't tracing, but also usable normally. It makes interpreting and debugging the interactive benchmark logs easier to see what happened and what didn't.
So making them verbose-only would be more reasonable, but since the rest are normally printed right now anyway, I suppose this can stay for now.