Add AMX with TensorFlow quick-start guides and AI use-case examples

[othakkar]

Adds documentation under software/tensorflow/ for leveraging Intel AMX acceleration with TensorFlow, including:

• Quick-start guide (README.md) — environment setup, BF16 matrix multiplication benchmark, and AMX verification via oneDNN verbose logging.
• NLP — BERT-Large (nlp-transformers-bert/) — sequence classification with Hugging Face Transformers, Keras mixed precision, TF Serving deployment, and REST client inference.
• GNN — RGAT (graph-neural-networks-rgat/) — relational graph attention network inference using TF-GNN with GATv2Conv, TF Serving deployment, and REST client inference.
• CV — ResNet50 v1.5 (computer-vision-resnet50/) — image classification with Keras built-in ResNet50, mixed precision, TF Serving deployment, and optional graph freezing.

Each guide covers Keras mixed_bfloat16 setup, model export, Docker-based TF Serving with --mixed_precision=bfloat16, client code, and AMX verification steps.

[rsiyer-intel]

Please update main README - https://github.com/intel/optimization-zone/blob/main/README.md to include a link to your contribution on TensorFlow in the software section.

[harp-intel]

Looking good. A couple requests for your verification. Also, don't forget to add a link in the repo's root README.md to the software/tensorflow/README.md.

[lucasmelogithub]

On the prerequisite for each README.md:

1. Is there a specific version of python we need to call out?
2. I suggest adding "docker" as a prerequisite

[othakkar]

Thank you for the review comments, @rsiyer-intel, @harp-intel, and @lucasmelogithub! I've addressed most of them. Regarding the Python version specification, this is typically handled implicitly through TensorFlow's dependency requirements, as each TF version enforces compatibility with specific Python versions.

[rsiyer-intel]

Thanks Om! Looks good.

[harp-intel]

Thanks @othakkar