Enable lazy loading support for DataCatalog 2.0 on Kedro-Viz

[SajidAlamQB]

Description

Related to: #2213

This PR integrates the new lazy-loading feature from Kedro’s updated DataCatalog API into Kedro-Viz.

Specifically, it checks if the project’s catalog is an instance of KedroDataCatalog. This allows Kedro-Viz to show dataset details without requiring full dataset installation. It also prevents unintended data loads if a dataset is configured to be lazy.

Development notes

• Conditional checks for KedroDataCatalog (DataCatalog v2) to enable lazy dataset behavior.
• Skips the “lite patch” if using the new catalog, because it’s not needed for lazy loading.
• Maintains backwards compatibility for older Kedro versions by falling back to the existing DataCatalog approach.

QA notes

Checklist

• Read the contributing guidelines
• Opened this PR as a 'Draft Pull Request' if it is work-in-progress
• Updated the documentation to reflect the code changes
• Added new entries to the RELEASE.md file
• Added tests to cover my changes

[ElenaKhaustova]

I think with the new catalog we can significantly simplify overall logic, which will be the same for default and lite modes:

1. We still need to resolve dataset factory patterns, but we do it via the catalog.config_resolver.resolve_pattern() method—it returns dataset configuration and does not require dataset initialization.
2. The above step is the only step needed to display the dataset
3. Once the user clicks on the dataset, we get it from the catalog, then call preview and process failure if something is uninstalled
4. That way, we don't really need UnavailableDataset and lite mode becomes default.

I see that the difficulty is in supporting different catalogs simultaneously and fitting different concepts together. Maybe we need to discuss how to apply the above concept to further transition to a new catalog while maintaining backward compatibility.

[ravi-kumar-pilla]

is kedro_obj not needed for TaskNode ?

[SajidAlamQB]

For the new catalogs, kedro_obj might be None for some nodes so we skip this. We store None for dataset_obj so we don’t cause an import or library load for the lazy loading approach.

[rashidakanchwala]

let's test this out in kedro viz --lite, remove kedro-datasets from pip. and see what happens.