init LOADER_IMAGE with env variable MODEL_LOADER_IMAGE

[googs1025]

What this PR does / why we need it

Which issue(s) this PR fixes

Fixes part of #350 (comment)

Special notes for your reviewer

Does this PR introduce a user-facing change?

init LOADER_IMAGE with env variable MODEL_LOADER_IMAGE in yaml file

[googs1025]

/kind cleanup

[googs1025]

Coming back here again, if we put this value in playground.backendConfig.env, it will be rendered into service and lws, and also rendered into pod env 🤔 , so it seems to be a convenient way to get this value. But I'm not sure if it's the right way this time.

like this:

func (p *ModelHubProvider) InjectModelLoader(template *corev1.PodTemplateSpec, index int) {
	initContainerName := MODEL_LOADER_CONTAINER_NAME
	if index != 0 {
		initContainerName += "-" + strconv.Itoa(index)
	}
	var loaderImage string
	for _, container := range template.Spec.Containers {
		for _, env := range container.Env {
			if env.Name == "MODEL_LOADER_IMAGE" {
				loaderImage = env.Value
				break
			}
		}
		if loaderImage != "" {
			break
		}
	}

	if loaderImage == "" {
		loaderImage = pkg.LOADER_IMAGE
	}

	// Handle initContainer.
	initContainer := &corev1.Container{
		Name:  initContainerName,
		Image: loaderImage,
		VolumeMounts: []corev1.VolumeMount{
			{
				Name:      MODEL_VOLUME_NAME,
				MountPath: CONTAINER_MODEL_PATH,
			},
		},
	}
        ....

}

[googs1025]

Currently the way this works is to get it from the controller env. I'm not sure if this is what we want.

[kerthcet]

Can we use configmap for long-term solution, because we may introduce more configures with ai gateway introduced, I think it doesn't require too much efforts.

[googs1025]

it means we don't need this change?

[kerthcet]

We need the feature but we need to use configmap to set the value instead.

[googs1025]

/close