Gemini model reasoning_effort in agy to use low, medium, high #38

I added it with this:

Overview of Changes

Added support for the --model flag and reasoning_effort mapping in the Google Antigravity CLI (agy) provider, mapped the corresponding Gemini 3.5 Flash reasoning models in the config, and improved configuration syncing on startup.

1. AGY Provider Update

• File updated: packages/server/src/providers/agy-provider.ts
• Details: Updated buildArgs to parse options.model and options.reasoningEffort. If a reasoning effort (e.g., low, medium, high) is specified, it is dynamically appended to the actual model name as a suffix (e.g., gemini-3.5-flash-low) and passed to agy via --model <model>.

2. Default Config Mappings

• Files updated: config.yaml & config.example.yaml
• Details: Added three new model mappings under model_mappings routing to agy to enable Gemini 3.5 Flash reasoning presets:
  • gemini-3.5-flash-low (actual model: gemini-3.5-flash, effort: low)
  • gemini-3.5-flash-medium (actual model: gemini-3.5-flash, effort: medium)
  • gemini-3.5-flash-high (actual model: gemini-3.5-flash, effort: high)

3. Startup Configuration Sync

• File updated: packages/server/src/db/seed.ts
• Details: Changed seedDatabase to upsert/sync config.yaml model mappings to the SQLite database on startup (rather than only seeding if the database is entirely empty). This allows new configuration file changes to propagate automatically on server restart.

4. Testing

• File updated: packages/server/src/providers/agy-provider.test.ts
• Details: Updated the provider unit tests to verify that --model is correctly populated with the dynamic reasoning effort suffix. All tests compiled and passed.