Faster processing of the exclusion list

[marcelm]

In summary, processing a long exclusion list (100000 entries) now takes a couple of seconds instead of days.

Previously, we would compare every detected clone id to every clone id in the exclusion list. Each comparison uses the is_similar() function, which implements the comparison by looking at each character of the clone IDs in turn and comparing them.

So if we have m detected clone IDs, n clone IDs in the exclusion list and clone IDs have a length of k, runtime would be O(mnk).

I initially thought that the comparison to each excluded clone ID is necessary because is_similar allows for approximate matches. However, it turns out that is_similar is called with max_hamming=0, that is, only exact matches are allowed anyway, so it is equivalent to just look up each detected clone ID in a set that contains the excluded clone IDs. Runtime becomes effectively O(m) (kind of, if we consider the C-level operations to take constant time).

[marcelm]

Merging this as it is an urgent fix. The test failure does not have to do anything with the change at hand, and we can take of it later.

[marcelm]

The test failure was relevant, there is actually a difference in results. Will investigate.

[marcelm]

The difference in results is alleviated by the additional changes in #72. (Which makes the processing of the exclusion list a bit slower again, but it is still just a couple of minutes.)