feat: Support model observability to pyfunc model

[tiopramayudi]

What this PR does / why we need it:

We introduce PyFuncV3Model as new pyfunc class that support model observability. It is working fine for new model but it is not the case for existing model that use PyFuncModel since merlin model can't be changed once it is logged or deployed. Implication of this, users need to create new model that use pyfunc_v3 type and the endpoint will be different hence there will be modification on the upstream caller. This PR will allow PyFuncModel to user model observability as long as the output of the model is PyFuncOutput.

This PR also allow features, entities, and prediction in PyFuncOutput to be pandas DataFrame, previously the type is strictly Values where in reality most of our users use pandas DataFrame and numpy ndarray

Modification:

• api/api/validator.go -> Relax the validation to allow PyFuncModel also
• api/cluster/resource/templater.go -> Supply necessary env vars for PyFunc and PyFuncV3 if model observability enabled
• python/sdk/merlin/pyfunc.py -> Supporting used of pandas DataFrame and numpy ndarray for model input and output

Which issue(s) this PR fixes:

Fixes #

Does this PR introduce a user-facing change?:

Checklist

• Added unit test, integration, and/or e2e tests
• Tested locally
• Updated documentation
• Update Swagger spec if the PR introduce API changes
• Regenerated Golang and Python client if the PR introduce API changes

[ghost]

👇 Click on the image for a new way to code review

Legend

[khorshuheng]

lgtm!

[krithika369]

@tiopramayudi @khorshuheng as discussed internally, let's explore moving the model type to the model version instead to solve the problem:

for existing model that use PyFuncModel since merlin model can't be changed once it is logged or deployed

Edit: Proceeding with this change at least for the short term, after internal discussions.

[krithika369]

Left one comment. The rest LGTM. Thanks, @tiopramayudi !