[WIP] Weighted random seed selection

[jiradeto]

@wuestholz, this PR adds new input selection mechanism that randomly picks inputs to fuzz based on the weight of inputs.

In AFL the fuzzed inputs are not saved in array but instead a linked list so there is no convenient way to maintain the index of inputs in the queue. This PR adds a new array that keeps track of all inputs so far in the queue.

[wuestholz]

@jiradeto Thanks a lot! I left some comments.

[jiradeto]

@wuestholz Thank you very much for your reviews. I addressed all the comments above and I think this PR is ready for evaluation. What do you think?

[wuestholz]

@jiradeto Great! Thanks! Yeah, lets give this a try. :)

[jiradeto]

@wuestholz, the fuzzbench result of the 20-hour fuzzing is now available:

• sqlite and bloaty: https://eloquent-bohr-97b0cb.netlify.app/results/2021-07-14-weight_random_selection/sqlite_bloaty/
• libjpeg and freetype: https://eloquent-bohr-97b0cb.netlify.app/results/2021-07-14-weight_random_selection/libjpeg_freetype

[jiradeto]

The next configuration:

• afl
• current version: 6dab941
• customized paramater version: decrease energy of the input that never been fuzzed

 if (!q->was_fuzzed) {
     w *= 1.0;
 }

[wuestholz]

@jiradeto Based on the experiments, it looks like 1.0 is better. Let's change that and add a comment that documents that we went from 5.0 (based on original probabilities in AFL) to 1.0.

[wuestholz]

@jiradeto Thanks! Could you please also apply your latest change to the combined PR (#6)?

[jiradeto]

Close this PR because it's implemented in #6.