Fix sub-slicing topology and level labels

[jamOne-]

Description

Changes:

• slice level label changed from google.com/gke-tpu-slice-{topology}-id -> cloud.google.com/gke-tpu-slice-{topology}-id
• removed 2x2 sub-slicing topology and set v6e-2x2 sub_slicing_available=False. Justification: subhost slices are not available and 2x2 is a subhost slice as each host has 2x4 chips.
• it turns out that Kueue TAS requires nodes to contain all Topology labels in order to schedule workloads, meaning we have to generate Topology CR dynamically during cluster create, based on the device type.

Issue

Testing

• Tested manually.
• Unit tests coverage.

[github-actions]

🤖 Hi @jamOne-, I've received your request, and I'm working on it now! You can track my progress in the logs for more details.