added more tests with hashCode=0 collisions

[jurgenvinju]

I'm debugging some issues with relation and map key removal on the vallang level and rascal level. They all relate to hashCode=0 collisions so I added some tests here to rule out a bug in capsule.

[jurgenvinju]

Here we see that after deleting one of the colliding nodes, the remaining node and the original non-colliding node have switched printing order:

equals fails then on:

if (rawMap1() != that.rawMap1()) {

in BitmapIndexedSetMultimapNode::equals

But I have yet to construct a test on the capsule level that reproduces this behavior. In Rascal:

rascal>{<"A",[]>, <"\a00",[]>, <""(),[]>} - {<"\a00",[]>} == {<"A", []>, <""(),[]>}
bool: false

• "\a00" and ""() both have hashCode zero
• "A" does not collide with the others
• it does not matter what the image values are with each key
• it does not matter which of the two colliding keys we remove
• if we remove the non-colliding key, they bug is not triggered

[msteindorfer]

The information provided in this PR, describing the observed behaviour, expedited triaging and fixing the underlying issue in #42.

Closing this PR in favour of #42, which suggests that there is already sufficient coverage in the property-based test suite to trigger the observed behaviour. After adding the invariant explicitly, the suite exhibited failures.