[Operators] Fix #43 - Improved exploit max length for dangling operator

[falcaopetri]

This PR implements the modification proposed in #43. I've updated all mining assistants that I could find in master.

This PR would also benefit assistants on other branches (such as the Injective assistant).

I've run some basic tests with -maxad 3 and -maxad 4, comparing mined rules before and after this patch. TSVRuleDiff reports no difference between mined rules.

This PR is slightly faster than master, and there is a huge performance improvement when mining constants. Also, AMIE queue calls to the last generation drops by about half.

Basic benchmark:

Mean of 5 runs in a 500k triples KB.

Without constants:

Summary
  'java -jar dangling_amie3.jar -bias lazy -oute -full -minhc .1 -maxad 3 kb.nt' ran
    1.02 ± 0.05 times faster than 'java -jar amie3.jar -bias lazy -oute -full -minhc .1 -maxad 3 kb.nt'

With constants:

Summary
  'java -jar dangling_amie3.jar -bias lazy -const -oute -full -minhc .1 -maxad 3 kb.nt' ran
    6.08 ± 0.21 times faster than 'java -jar amie3.jar -bias lazy -const -oute -full -minhc .1 -maxad 3 kb.nt'

[lgalarra]

I agree with these changes. They make the logic neater. @lajus since it is a critical piece of code, I will feel more relieved if you also approve it.