Optimizing weights through the MAB algorithm

Hello, I've noticed that in Fuzzilli, Mutators, CodeGenerators, and Templates all use fixed weights for selection, which I believe is not suitable.
Here are some experiments I've done: the probability of each Mutator generating an interesting program changes over time.

It can be observed that the CombineMutator and SpliceMutator have the highest interesting rates, but these two mutators are not assigned higher weights. Therefore, I want to optimize the weights of the mutators through the MAB algorithm. I've implemented the related code locally. In short:

• Selecting a Mutator is considered as a pull. If the Mutator generates an interesting program, it is considered a profit.
• The Thompson sampling algorithm is used to estimate the Mutator with the highest interesting rate and select it more times.
• A MABList is created to replace the original WeightedList. The MABList will automatically select the appropriate elements based on feedback.

I would like to know if you would accept this PR if I were to propose it?

The issue proposes replacing WeightedList with MABList for mutators, code generators, and templates using Thompson sampling. Start by tracing how these elements are selected and how interesting-program feedback is reported. Done would require maintainer agreement on the design and evidence that adaptive weighting works without disrupting existing fuzzing behavior.