Regarding performance

@zvr After some profiling, turns out a massive performance hit occurs due to our "file entries re-usage".

While compiling the xbps package manager, i made the following measurements:

(All in CPU time)

without build recorder wrapping it: 6.934s with build recorder wrapping it: 14.988s

That's an overhead of 8.064s, of which 5.180s is due to get_file_hash function call at src/tracer.c:255.

This raises the question, do we really need to search using the hash as well? Absolute paths are unique to begin with and we are already tracking rename(2) calls. This paired with the fact that the finfo array is sorted in ascending order with respect to time, is a guarantee that the first match on a backward search is the file we are looking for.

Is my logic flawed? Is the hash check there important?