Fix test_illegal_put failure in plasma test.

[robertnishihara]

In this test, client1 and client2 are connected to manager1 and manager2 respectively. We call client1 calls create and seal on an object. Then client2 calls create and seal on a slightly different object with the same object ID. This should cause a hash mismatch followed by a fatal error. We then check that the plasma manager died. However, in many instances on Travis, this check failed even though we could see the fatal error in the logs.

I think the issue is that manager1 was actually the one dying in a bunch of cases, meaning that the timing was such that the RAY.OBJECT_TABLE_ADD command for the second manager arrived at Redis first. This seems pretty unlikely, but I guess it can happen.

[AmplabJenkins]

Merged build finished. Test PASSed.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/40/
Test PASSed.

[AmplabJenkins]

Merged build finished. Test PASSed.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/41/
Test PASSed.