Closesly modelled after the tfevent-metricscollector.
Currently not yet working, as there are issues
that the arguments from the `injector_webhoook`
are somehow not passed.

Addresses: kubeflow#2019

The TrialName can be parse from the pod name.

This seems currently a good way to get the trial name. For
more discussion see: kubeflow#2109

This example illustrates how a full kfp pipeline can
be tuned using Katib.

It is based on a metrics collector to collect kubeflow
pipeline metrics (kubeflow#2019). This is used as a Custom Collector.

Addresses: kubeflow#1914, kubeflow#2019

Before the notebook only worked with Python 3.11.
Now it is also tested with 3.10

Also the experiment/run name is extended with a timestamp
for easier reruns.

Otherwise the image was binarized, leading to an
artifically bad performance.

And remove an old comment

Co-authored-by: axel7083 <42176370+axel7083@users.noreply.github.com>

As per suggestion

As suggested in the PR review, the generic case where multiple KFP pipeline
metrics files would be present in the output folder is supported.

Note that in the current KFP v1 implementation always only one data
file is present.

As per suggestion this should make it easier to handle the v2 metrics
collector in the future as well

This installs Kubeflow pipelines (KFP) if selected to do so in order to
run e2e tests where Katib and KFP interact.

This commit should be removed later

These permissions are required such that the katib-controller can launch
argo workflows.

This adds a dummy e2e example that can be used to test the main
functionality.

This reverts commit 36ed372.

This could be used for e2e testing

Otherwise the patching of the `katib-controller` cluster role would
not work.

This enables the user to set th version of the KFP
version which should be useful to use this script
to install KFP v1 and v2 without additional parameters.

This is required for kubeflow pipelines as I found no easy way
to install kubeflow pipelines into the `default` workspace
that was previously the hardcoded one.

Now the namespace can be passed as a parameter.

This action should now run the kubeflow pipeline v1 e2e example.

This required the extension of the `template-e2e-test` to include
parameters to
a) install kfp
b) select the `kubeflow` namespace (instead of default) to run the tests
with.