* Test reliability fix

There were some issues with the original test:

- Config changes to riak_core where put under default bucket properties, and so were not applied.  In particular, although the test was run 3 times on different ring sizes, only the default ring size (64) was actually tested.

- In testing the genuine ring sizes, it was uncovered that the Ring9 test will fail on ring sizes other than 64.

- As the test did not wait for transfers, they were not a genuine system test, and there was potential for false success and false failure due to this.  The test now waits for transfers to complete; but does so in a fairly cumbersome way as `rt:wait_until_transfers_complete/1` is not a reliable function (in particular when it is called before any transfers have started.

- After adding the wait for transfers, and correcting the vnode_inactivity_timeout, it was uncovered that there exists a race condition whereby a vnode may be simultaneously be exiting (triggered by a join) and be required (to receive a handoff).  This could lead to transfers that got permanently stuck due to `Transfers: [{stopped,'[email protected]',6}]`.  A timeout of 60s is sufficient to make the test relatively reliable, but significantly extends the time required to run the test.

* Update location.erl

* Add leave test

* Update leave test

* Update core_all