Enhance hard code for export dir in deploy component

[jinchihe]

While using deployer component to create a TF-Serving service, hit a issue that could not find base path for servable. Checked the details, that's caused by the export directory is hard code to model_path/export/export as below in deploy.sh.

ks param set server modelPath "${MODEL_PATH}/export/export"

To let the deployer component to be reusable, we should enhance that because the model export path can be specified during tensorflow/estimator training from the offical API.

For the TFX taxi sample, the export path is specified model_path/export/export in training step, so update taxi-cab-classification-pipeline.py as well. Thanks.

---

This change is 

[jinchihe]

/assign @gaoning777

Could you please review the ticket? Thanks.

[gaoning777]

Another place that needs modification is the notebook sample: https://github.com/kubeflow/pipelines/blob/master/samples/notebooks/KubeFlow%20Pipeline%20Using%20TFX%20OSS%20Components.ipynb.
Please update the deployer component in this sample, too.

[jinchihe]

Another place that needs modification is the notebook sample: https://github.com/kubeflow/pipelines/blob/master/samples/notebooks/KubeFlow%20Pipeline%20Using%20TFX%20OSS%20Components.ipynb.
Please update the deployer component in this sample, too.

Addressed you comments. Great thanks for point out the missing. @gaoning777

[gaoning777]

/test kubeflow-pipeline-sample-test

[jinchihe]

@gaoning777 Any more comments? Thanks for reviewing.

[gaoning777]

I saw the sample test failure. Do you mind taking a look at this?
The sample test is not explicitly shown in the github but we enabled them in the background.

https://gubernator.k8s.io/build/kubernetes-jenkins/pr-logs/pull/kubeflow_pipelines/823/kubeflow-pipeline-sample-test/736/

[gaoning777]

Never mind. Since the new deployer component has not been built, thus sample failures with --model-export-path not found. This PR will lead to samples failures.

Could we separate this PR into two:

1. deployer code change.
2. sample change.
   When this first PR is merged, I will release the component images.
   In the second PR, I can also help update the image tags in the samples.
   This way, we will not have broken samples during any time.

[jinchihe]

Never mind. Since the new deployer component has not been built, thus sample failures with --model-export-path not found. This PR will lead to samples failures.

Could we separate this PR into two:

1. deployer code change.
2. sample change.
   When this first PR is merged, I will release the component images.
   In the second PR, I can also help update the image tags in the samples.
   This way, we will not have broken samples during any time.

@gaoning777 Thanks for your comments!
I updated the PR to change the deploy script, and created new one #843 to update the sample TFX taxi.

[gaoning777]

/lgtm

[gaoning777]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: gaoning777

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:

• components/OWNERS [gaoning777]

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