fix consistent clearing of internal cache entries during shutdown

[zyclonite]

should fix #92 and #94

[zyclonite]

@tsegismont i honestly do not like the vertx.close() either... but isn't that triggering a graceful shutdown and stopping of all verticles anyway?

the reason for doing the stop is, that ignite stops the member node and so all tries to access the cache will fail from that point on (so there seems to be no reason to keep anything alive) this happens only on split-brain situations or when all the configured timeouts are passed (depends as well on the base topology)

[tsegismont]

@zyclonite I understand the logic behind this line. But I fear this would cause more problems than it solves.

Perhaps this could be made configurable?

[zyclonite]

@tsegismont sure i can do that

[zyclonite]

@tsegismont from my perspective this can be merged