Hide extra dependencies from boosting mutation

Solution is to make boosting operation a callable class instead of just a function. This will help hiding extra dependencies in `__init__` of the new class:
- drop this access to `advisor` in the line: `task_type = params.advisor.task.task_type`
- repeated calls to `OperationTypesRepository` that return the same results

boosting mutation will be built once and provided to `MutationOperator`