Record statement copies during loop unrolling

[sim642]

This resurrects my old attempt at identifying syntactically unrolled loop iterations for #1372.
The key yet subtle part to making it work was to only use skind and labels for stmt hashes for the hashtable.

This also adds a CFG cram test where statement copies are indicated by dotted edges.

TODO

• Test nested loops.

[michael-schwarz]

There seems to be some issues with the OS X cram tests:

• cfgDot: command not found
• graph-easy: command not found
• goblint --set lib.activated '[]' --set exp.unrolling-factor 5 --enable ana.int.interval --enable witness.yaml.enabled 11-unrolled-loop-invariant.c seems to have radically different output than what was expected