[ENHANCEMENT] Auto-refresh ClusterBaseModel config.json upon operator update to populate missing fields

[YouNeedCryDear]

What would you like to be added?

When the OME operator updates, runtimes might expect new fields in the ClusterBaseModel (such as quantization). Old versions of ClusterBaseModel created with an older model agent will be missing these fields, making them incompatible with the new runtime under the new OME version.

We need to introduce a controller logic to:

1. Detect missing required fields in existing ClusterBaseModel resources.
2. Automatically refresh the model metadata by re-parsing the config.json file to populate the missing fields, without triggering a full re-download of model weights.

Why is this needed?

This ensures backward compatibility and smooth upgrades. Without this, old ClusterBaseModel resources break upon upgrading the OME version because the runtimes fail to start without the newly expected fields.

Acceptance criteria

• The ClusterBaseModel controller detects when existing resources are missing fields expected by the new operator version.
• A metadata refresh process is triggered for these resources.
• The refresh logic re-parses the existing config.json file and successfully populates the missing fields.
• The model weights are NOT re-downloaded during this metadata refresh.
• Runtimes using the updated ClusterBaseModel start successfully.

Completion requirements

• API change
• Docs update
• Tests

Can you help us implement this enhancement?

• Yes, I can contribute
• No, but I'm available for testing
• No