[CI] Try to stabilize Travis job (#685)

* Try to stabilize Travis job

There are some cases where Travis rarely fails due to an unknown error.
I think we need a little bit of waiting time to finish one test, delete the clipper cluster,
and then recreate the clipper cluster for a new test. Let's try one patch to solve this problem.

* Add python(2.7) to travis.yml

Travis shows this message now.
'This job is running using the default Python version, which was updated to 3.6 on April 16th, 2019. You can explicitly stay on the previous version by specifying python: 2.7 in your .travis.yml.'

.travis.yml

@@ -10,6 +10,8 @@ sudo: required
 
 language: python
 
+python: "2.7"
+
 env:
   matrix:
   # This will make sure .minikube config resides in current user folder

bin/run_travis.sh

@@ -68,9 +68,9 @@ retry_test() {
 
 # if the test test succeed, debug info will not get printed
 # mainly used to debug container being evicted
-retry_test python kubernetes_integration_test.py
-retry_test python kubernetes_multi_frontend.py
-retry_test python kubernetes_namespace.py
+retry_test python kubernetes_integration_test.py; sleep 30
+retry_test python kubernetes_multi_frontend.py; sleep 30
+retry_test python kubernetes_namespace.py; sleep 30
 retry_test python multi_tenancy_test.py --kubernetes
 
 # TODO: disabled for now, will re-enable after RBAC PR