Add Trial images build to the CI

[andreyvelich]

Blocked by: kubeflow/testing#923.

I added build of Katib Trial template images to the CI to make sure that each Trial template example can be built and executed.

As well, I fixed few problems with the current images:

• Fix Downloading MNIST dataset with torchvision gives HTTP Error 403 pytorch/vision#1938 problem in PyTorch mnist example.
• Add --log-path flag to PyTorch mnist to use this example in File and StdOut metrics collector mode.
• Use docker.io/kubeflowkatib/pytorch-mnist instead of gcr.io/kubeflow-ci/pytorch-dist-mnist-test:v1.0 since it is similar images.
• Modify PyTorch mnist Dockerfile to be consistent with MXNet mnist
• Use mxnet/python:latest_cpu_native_py3 base image in MXNet since it has all required MXNet setup.
• Use tensorflow/tensorflow:1.15.4-py3 and tensorflow/tensorflow:1.15.4-gpu-py3 for ENAS CNN Trial template. With that setup we can have only 1 requirements.txt for this image and we can cut the Dockerfile.

After we merge this PR, I will execute release script to push new images to the kubeflowkatib registry. Then I will update tags for the Katib Trial templates.

/assign @gaocegege @johnugeorge

/cc @PatrickXYS

[google-oss-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: andreyvelich

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [andreyvelich]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[andreyvelich]

@gaocegege @johnugeorge I think the tests are working fine, can you take a look please ?

[gaocegege]

/lgtm