[Core] Refactor quantization parsing into base model config

[YouNeedCryDear]

What this PR does

• moves quantization parsing into BaseModelConfig
• removes duplicated quant parsing from individual model config implementations
• updates phi to rely on the shared base model config behavior

Why we need it

The HuggingFaceModel interface requires every model configuration object to implement the GetQuantizationType() string method. Currently, over 30 distinct architectural implementations (e.g., QwenConfig, LlamaConfig, DeepseekV3Config) successfully satisfy this requirement.

However, almost none of these implementations utilize a centralized method. As a result, 90% of the model implementation files contain identical copy-pasted blocks redefining the JSON configuration struct and the accessor method:

This consolidates quantization parsing in one place so model config implementations stay consistent and easier to maintain. It also removes duplicated parsing paths that had drifted across individual model types.

Fixes #

How to test

• Run go test ./pkg/hfutil/modelconfig
• Optionally run go test ./pkg/hfutil/... in an environment that allows httptest to bind local ports for pkg/hfutil/hub tests

Checklist

• Tests added/updated (if applicable)
• Docs updated (if applicable)
• make test passes locally

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!