Fix Unintended Behavior of Solution Selection in Training Configuration

[Granoffel]

Reason (Why?)
In our current training configuration interface, we have observed unexpected behavior related to the selection and visibility of machine learning (ML) libraries and AutoML solutions. Specifically, users are able to deselect all ML libraries, which contradicts the intended functionality that requires at least one library to be selected at all times. Additionally, when deselecting ML libraries, previously hidden AutoML solutions become visible again. This issue disrupts user experience and can lead to confusion during the configuration process, affecting the efficiency and effectiveness of setting up training environments.

Solution (What?)
The proposed solution involves implementing validation logic to ensure that at least one ML library remains selected at any given time. This can be achieved by disabling the de-selection of the last remaining selected library or by automatically re-selecting a default library if a user attempts to deselect all. Furthermore, the visibility state of AutoML solutions should be correctly tied to the selection state of ML libraries, ensuring that deselected libraries do not inadvertently trigger the visibility of any AutoML solutions that were intentionally hidden. These adjustments will streamline the configuration process and ensure a more logical and user-friendly interface.

Acceptance criteria
It is impossible for a user to deselect all ML libraries; attempting to do so will either not deselect the last selected library or will automatically re-select a default library.
Deselecting ML libraries does not cause previously hidden AutoML solutions to become visible.