Add support for Bitbucket Server and integrate LiteLLM for better LLM interoperability

[JimiHFord]

🚀 Multi-SCM Provider Support & LiteLLM Integration

Summary

This PR introduces two major enhancements:

• Multi-SCM Provider Support: Adds Bitbucket Server alongside existing GitLab integration using a strategy pattern.
• LiteLLM Integration: Replaces direct pydantic-ai provider usage with LiteLLM for unified LLM provider support (100+ providers).

New Features

🔌 SCM Provider Abstraction

• Strategy pattern implementation with abstract SCMProvider base class.
• GitLab provider: Refactored existing GitLab code into a dedicated provider.
• Bitbucket Server provider: New implementation using atlassian-python-api.
• Provider-agnostic terminology: Repository (not project), Pull Request (not MR), Namespace (not group).
• Factory pattern for provider instantiation based on SCM_PROVIDER environment variable.

🤖 LiteLLM Integration

• Unified LLM interface supporting 100+ providers (OpenAI, Azure, Anthropic, Bedrock, Vertex, Ollama, etc.).
• Simplified configuration: Provider inferred from model name prefix (e.g., azure/gpt-4o, anthropic/claude-3-opus).
• Automatic quirk handling: LiteLLM manages provider-specific differences like max_tokens vs. max_completion_tokens.
• Built-in retry support: Removed custom retry_client.py in favor of LiteLLM's native retry handling (num_retries parameter). LiteLLM automatically handles rate limits (429), server errors (500/502/503/504), and respects Retry-After headers.

Files Changed

Category	Files
New SCM Providers	scm_providers/ (base, factory, gitlab, bitbucket_server)
New LLM Module	llm/ (litellm_model, factory)
Updated Agents	analyzer.py, documenter.py, ai_rules_generator.py
Updated Config	config.py, .env.sample
Refactored	handlers/cronjob.py (now uses SCM interface)
Removed	utils/retry_client.py (replaced by LiteLLM retries)

Configuration

SCM Provider Selection

SCM_PROVIDER=gitlab  # or bitbucket_server

LLM Provider Examples

# Each agent can use a different provider
ANALYZER_LLM_MODEL=azure/gpt-4o-deployment    # Uses AZURE_* env vars
DOCUMENTER_LLM_MODEL=claude-sonnet-4-20250514 # Uses ANTHROPIC_API_KEY
AI_RULES_LLM_MODEL=gpt-4o                     # Uses OPENAI_API_KEY

Deprecations

• --group-project-id CLI arg deprecated in favor of --namespace-id for provider-agnostic terminology. Still works but emits deprecation warning. Will be removed in v2.0.

Dependencies Added

• litellm>=1.55.0: Unified LLM provider interface.
• atlassian-python-api>=3.41.0: Bitbucket Server API client.

Testing

• ✅ Analysis tested with Azure OpenAI via LiteLLM.
• ✅ All 5 analyzer agents completing successfully.
• ✅ Linting passes (ruff check src/).
• ✅ Deprecation warning verified for --group-project-id