WIP: Pass RHELAI_IMAGE to pytorchjobs via envvar

[gmfrasca]

Pass the pytorchjob base image reference via environment variable

Description

Get around hardcoded RHELAI Image used for Training PyTorchJobs by passing the const (which is already defined in util/consts.py) to the manifest op via an env var. This will mean that all image references are now centralized in util/consts.py

How Has This Been Tested?

Merge criteria:

• The commits are squashed in a cohesive manner and have meaningful messages.
• Testing instructions have been added in the PR body (for PRs involving changes that are not immediately obvious).
• The developer has manually tested the changes and verified that the changes work

[gmfrasca]

This would supersede #248 .

I think this is a bit cleaner, albeit less direct, way to get the RHELAI_IMAGE ref to the manifest op rather than plumbing the value directly through via component and pipeline params

[leseb]

I believe #248 was more straightforward. Using an environment variable feels more like a workaround.

From my perspective, components should be self-contained and “portable.” Introducing this environment variable as a dependency creates a hidden requirement that isn’t explicitly part of the function definition. If we ever need to move this component to another pipeline, we’ll have to set the environment variable again. In contrast, having all the required values defined within the function itself makes the dependency much clearer.