Extend specs of Set#divide

[herwinw]

This contains two commits:

1. Add test for calling Set#divide without a block. There was a commented piece of code that expected an exception, but since Ruby 2.4 the documentation states it should return an Enumerator
2. Add test for calling the block with an undefined arity. The docs describe the behaviour for blocks with arity 1 and 2, but not what happens when call the method with a block with a different arity. This is technically undefined behaviour, so I'm not sure whether this spec should be added.

[andrykonchin]

suggestion: I would move out this case into a separate describe/`context' (with a title like "when passed a block with an arity > 2") for consistency.

[andrykonchin]

suggestion: Also I like the check in the case above:

ret.each(&:even?).should == Set[Set[1, 3], Set[2, 4]]

using Set instead of sorting looks more readable and shorted

[andrykonchin]

suggestion: TBH I would find a case title a bit laconic and not comprehensive (as well as in a similar case below for arity = 2). It will be great if the titles reflect documentation and describe exactly what happens and how a set elements are compared/divided:

If the arity of the block is 2, elements o1 and o2 are in common if block.call(o1, o2) is true. Otherwise, elements o1 and o2 are in common if block.call(o1) == block.call(o2).

[herwinw]

suggestion: I would move out this case into a separate describe/`context' (with a title like "when passed a block with an arity > 2") for consistency.

I extracted this one into its own block, and added a block with a splat argument as well (which also should use the first element, since this results in an arity of -1, this might be a nice extra test case for Ruby implementations)

I think the other two entries are nice additions, but slightly out of scope for this case.

[andrykonchin]

Thank you for the specs!