feat: add configurable AI provider support

[d-claro]

Description

Added support for configurable AI providers so each project or team member can choose which AI service to use, without changing code.

Type of Change

• New feature

What was done

• Added ai_provider, ai_model and api_key fields to config.json
• Created load_config() to read configuration with safe defaults
• Created ask_ai() as a single entry point that routes to the configured provider
• Refactored ask and generate-test to use the new layer
• Both commands now show which provider and model is being used
• Claude and OpenAI show a clear "not enabled yet" message with instructions

Test Cases Affected

None — this change affects AI commands only.

How to Test

1. Run python agent/cli.py ask "What test cases do we have?" — confirm it works with ollama
2. Change ai_provider to claude in config.json
3. Run the same command — confirm a clear message appears explaining Claude is not enabled
4. Change back to ollama

Checklist

• I have tested this locally
• I have updated the Knowledge Base if needed
• I have assigned a reviewer

Notes

Claude and OpenAI integrations require API keys (separate from the Claude Teams subscription). The structure is ready — enabling them is a matter of filling in the corresponding block in ask_ai().