Add libFuzzer dictionary support

[attekett]

At the end of each run, libFuzzer prints recommended dictionary it has collected during the run.

Example:(libxml2 stub run.)

54598 DONE cov: 13393 bits: 76413 indir: 375 units: 5746 exec/s: 540

Recommended dictionary.

"id" # Uses: 2307
"foo:" # Uses: 1868

End of recommended dictionary.

Done 54598 runs in 101 second(s)

libFuzzer can use these dictionaries to improve its performance in the following runs.

We should collect these recommended dictionaries from each run and incrementally build larger dictionary that is fed as an input for the next fuzzing round.

We could also write the dictionary out into /results/.dict, so when we start fuzzing next time we could load dictionary built on previous fuzzing campaigns.