Make `foldl/4` error outside its domain

[mb64]

Use single-sided unification rules for foldl/4 so that something like foldl(not_a_list, Goal, In, Out) is an exception instead of silently failing. This would be a breaking change.

Some unresolved questions:

• Is it a good change? I had to debug some code today where this would have told me the problem right away. However, it is a breaking change, so there is a chance that someone relies on this behavior.
• foldl/N where N > 4? I originally also wanted to change the other foldls, but decided against it, since it would break all the code that uses foldl like a kind of "maplist with state".
• scanl? I think this change might be significantly more breaking for scanl than for foldl, so I didn't change scanl.
• Unit tests: I didn't add any, since I'm not familiar with the test setup, but with guidance I would be happy to add one.

[JanWielemaker]

Thanks. This breaks test src/Tests/library/test_apply.pl:117 :cry: We could argue this to be a buggy test. We could also argue that non-equal length lists for foldl/5+ should return an error. I'm not sure. Maybe you would like to discuss this on Discourse?