Update tensorflow notebook version to v20180403-1f854c44 (kubeflow#589)…

… (kubeflow#590)

Includes the following changes:
* Fix the tensorflow version in prebuilt images for python 2.7. (kubeflow#580)
* Install graphviz in tensorflow notebook image (kubeflow#583)

Update tensorflow notebook version to v20180403-1f854c44 (kubeflow#589)…

… (kubeflow#590)

Includes the following changes:
* Fix the tensorflow version in prebuilt images for python 2.7. (kubeflow#580)
* Install graphviz in tensorflow notebook image (kubeflow#583)

Cherrypick kubeflow#570 and kubeflow#572 into v0.1-branch (kubeflow#575)

* Update tf jupter notebook images to include tfma (from kubeflow#544) (kubeflow#570)

* Update tf jupter notebook images to include tfma (from kubeflow#544)

tf1.4 gpu image has issues building with the latest tfma, so that's
not updated in this PR

/cc @jlewi
/cc @ojarjur

* Don't update tf images older than 1.6

* Moved OAuth secret from param to named secret (kubeflow#572)

Add TfJob dashboard to ksonnet (kubeflow#548) (kubeflow#555)

Cherrypick kubeflow#526 and kubeflow#533 into v0.1-branch (kubeflow#540)

* Fix typos in tensorflow notebook images in spawner

* Update the TFJob image in preparation for a new RC. (kubeflow#533)

* The updated TFJob includes the fix to delete pods when job finishes
  rather than leave them running.

* We need to pin the commit of the tf-operator in the extra repos so
  that we checkout the tests at the commit that matches the image.

Update release instructions to ease onboarding (kubeflow#493)

* Work in progress on updating release instructions.

This commit:

1.  Adds instructions for authenticating to GCP,
2.  adds instructions for setting the right kubectl context, and
3.  ensures that workflow names are less than 63 characters.

* This commit improves instructions and fixes a spurious environment variable