refactor(llm): extract LLM client from main loop to improve modularity

[Lochit-Vinay]

Closes #468

🧩 Extract LLM client from main loop

Summary

This PR extracts the LLM API interaction logic from LLM.main_loop() into a separate module (llm_client.py).

The goal is to reduce tight coupling in the LLM pipeline and make future changes (validation, fallback handling, prompt improvements) easier to implement independently.

---

Changes

• Moved Ollama API call logic into src/llm_client.py
• Replaced inline request logic in llm.py with a function call
• No changes to existing behavior

---

Why this matters

Currently, multiple responsibilities (prompting, API calls, parsing, validation) are tightly coupled in the same flow.

This extraction:

• improves separation of concerns
• reduces overlap between feature PRs
• makes the pipeline easier to extend

---

Scope

• Refactor only (no functional changes)
• Existing behavior is preserved

---

Context

Part of addressing: #468

This is the first step in a series of small refactors to modularize the LLM pipeline.