Add testlib case for HashMap untreeifying during iteration

[ben-manes]

A really interesting edge case for HashMap was reported due to my usage, see ben-manes/caffeine#176.

It seems that when the entrySet iterator removes entries, the untreeifying causes the entries to be rewritten. Then the entries returned by the iterator are stale and setValue has no effect on the HashMap. This is because unlike concurrent maps, the setValue does not perform a map.put(key, value) to handle the race.

I believe this is a JDK bug and we are awaiting feedback. Guava's testlib doesn't assert this case, so it might be a good addition.

[ben-manes]

https://bugs.openjdk.java.net/browse/JDK-8186171

[kluever]

Sounds like something @lowasser might be interested in taking a look at.

[jbduncan]

@ben-manes The way I see it, it's still worth keeping this open since it seems to be interesting edge case worth testing against, at least for custom Map impls like ImmutableMap.

[ben-manes]

Sure thing. JDK bug was fixed, fwiw.