Initial Chat Template Implementation

[sayanshaw24]

Feature

Adds HuggingFace-like Chat Template support into ORT Extensions for the ONNXRuntime workload for Phi, Llama and DeepSeek models.

Validation

Parity with HF chat template output for the following models:

• Phi: microsoft/Phi-4-multimodal-instruct, microsoft/Phi-3.5-mini-instruct, microsoft/Phi-3.5-MoE-instruct, microsoft/Phi-3.5-vision-instruct, microsoft/Phi-3-vision-128k-instruct, microsoft/Phi-3-mini-4k-instruct, microsoft/Phi-3-mini-128k-instruct, microsoft/Phi-3-small-4k-instruct, microsoft/Phi-3-small-128k-instruct, microsoft/Phi-3-medium-4k-instruct, microsoft/Phi-3-medium-128k-instruct
• Llama: meta-llama/Llama-2-7b-chat-hf, meta-llama/Meta-Llama-3-8B-Instruct, meta-llama/Meta-Llama-3-70B-Instruct, meta-llama/Llama-3.1-8B-Instruct, meta-llama/Llama-3.1-70B-Instruct, meta-llama/Llama-3.2-1B-Instruct, meta-llama/Llama-3.2-3B-Instruct, meta-llama/Llama-3.3-70B-Instruct
• DeepSeek: deepseek-ai/DeepSeek-R1-Distill-Llama-70B, deepseek-ai/DeepSeek-R1-Distill-Qwen-32B

[natke]

Checking exactly for the template string seems a bit brittle - can we use model type instead?