Bug: discovery uses incorrect Intel resource & label names

[eero-t]

WVA device discovery and kind simulation use incorrect Intel extended resource & label names, and mismatching product names:

• intel.com/gpu.product + intel.com/gpu.memory GPU node labels
• Intel-Gaudi-2-96GB product name for a GPU resource
• intel.com/gpu GPU resource

Whereas the extended resource names provided by Intel device plugins are actually following:

• habana.ai/gaudi for Gaudi AI products
• gpu.intel.com/i915 for legacy GPU products
• gpu.intel.com/xe for recent GPU products

And node labels they provides are of the form:

• habana.ai/product.name
• gpu.intel.com/product

There's currently no automatic node labeling for device memory amounts, so those would need to be set manually, using the same pattern as rest of their labels:

• habana.ai/device.memory
• gpu.intel.com/memory

References:

• https://intel.github.io/intel-device-plugins-for-kubernetes/cmd/gpu_plugin/README.html#introduction
• https://intel.github.io/intel-device-plugins-for-kubernetes/cmd/gpu_plugin/labels.html
• https://github.com/HabanaAI/gaudi-device-plugin/blob/main/examples/hl-smi-test.yml#L29
• https://github.com/HabanaAI/gaudi-feature-discovery/blob/main/collector/collector.go

(Faulty info seems to originate at least partially from #580?)