Add gemma family for google provider

[khda-tech]

Add Gemma 3 Family Models for Google Provider

Summary

Added configuration files for the new Gemma 3 and Gemma 3n model families to the Google provider.

Models Added

Model File	Model Name	Parameters	Context	Output	Multimodal
gemma-3-4b-it.toml	Gemma 3 4B	4B	32K	8K	✅ Text + Image
gemma-3-12b-it.toml	Gemma 3 12B	12B	32K	8K	✅ Text + Image
gemma-3-27b-it.toml	Gemma 3 27B	27B	128K	8K	✅ Text + Image
gemma-3n-e2b-it.toml	Gemma 3n 2B	2B	8K	2K	❌ Text only
gemma-3n-e4b-it.toml	Gemma 3n 4B	4B	8K	2K	❌ Text only

Key Features Enabled

• ✅ attachment: All models support file/image attachments (where applicable)
• ✅ structured_output: Supported on 12B and 27B variants
• ✅ tool_call: Enabled for Gemma 3 27B
• ✅ open_weights: All models marked as open weights
• ✅ temperature: Configurable sampling supported

Sources

• Model metadata sourced from official documentation: https://ai.google.dev/
• Verified against live API response:

  curl "https://generativelanguage.googleapis.com/v1beta/models" \
    -H "Content-Type: application/json" \
    -H "x-goog-api-key: $GOOGLE_API_KEY" | jq '.models[].name'

Validation

• ✅ TOML syntax validated
• ✅ Field consistency checked against provider schema
• ✅ Model names match API identifiers (models/gemma-3-*-it, models/gemma-3n-*-it)
• ✅ Cost fields set to 0 (free tier / open weights)
• ✅ bun validate passes with no errors
• ✅ Verify models render correctly on models.dev