Discrepancy between setting save_version via catalog constructor and when passing datasets

[ElenaKhaustova]

Description

In the catalog constructor

kedro/kedro/io/kedro_data_catalog.py

Line 65 in 9464dc7

save_version: str | None = None,

and from_config

kedro/kedro/io/kedro_data_catalog.py

Line 273 in 9464dc7

save_version: str | None = None,

method we require passing one save version for all datasets. We further use this save version to instantiate all datasets objects.

This constraint is bypassed when passing instantiated dataset objects to catalog constructor.

Context

1. It's not clear whether this constraint is a requirement.
2. It's not clear what save version should we save and load back when serialization/deserialization [DataCatalog]: Spike - Catalog serialization and deserialization support #3932

Possible Implementation

1. Clarify the requirement of having single load_version per catalog.
2. Based on the first point either remove the requirement or apply it for instantiated datasets passed to the catalog.

Possible Alternatives