Add Ollama provider support for LLM and chat assistant flows

[pPyrius]

Summary

Adds first-class Ollama support across the FinBot LLM stack.

Changes

• Register ollama as a supported LLM_PROVIDER
• Add OLLAMA_MODEL config, used only when LLM_PROVIDER=ollama
• Add Ollama Python SDK dependency
• Route chat assistant requests through the configured LLM provider instead of always using OpenAI
• Convert OpenAI-style tool definitions/history into Ollama-compatible chat/tool messages
• Auto-pull the configured Ollama model on first use if it is not installed locally
• Update README and .env.example with Ollama configuration
• Add regression tests for Ollama routing, model selection, model pull behavior, and chat provider routing

Testing

• git diff --check
• python -m py_compile finbot/agents/chat.py finbot/core/llm/client.py finbot/core/llm/ollama_client.py finbot/config.py
• uv lock --check
• env DEBUG=true .venv/bin/pytest tests/unit/llm/test_llm_client.py tests/unit/llm/test_ollama_client.py tests/unit/agents/test_chat_provider_routing.py -q

Result:

45 passed, 2 skipped