bugfix in action predicate removal

[PLauerRocks]

Hi, this is my proposed fix for the action predicate removal issue.

I am not entirely sure it is correct. In particular I am concerned about the following two things:

• My implementation assumes, that if the action_predicate atom occurs in a body, it is always listed first. This seems reasonable to me, but I am not 100% certain I can assume that.
• In your previous implementation you did not change the weight of the rule if the condition size is != 1, this is different now. If this was important, then this needs to be adjusted in my implementation.

[abcorrea]

Thanks!

But if I understand the issue correctly, the problem is that the equality constraint shouldn't even be there: one doesn't need to enforce all predicates to occur at most once in the head.

This seems like a legacy from Fast Downward's translator codebase. I will have a look at that in the next few days, so maybe we can make the code more robust --- assuming you want to run it with gringo or idlv as a grounder but not with the Fast Downward one, is this right?

Regarding the weight: the change should probably go the other direction: remove the weight attribute as we never exploited that at end. Or do you use it for something, @PLauerRocks?