feat: Add native Google Gemini provider support

[mukund-gohil-atos]

Summary

CodeWiki currently supports openai-compatible, anthropic, bedrock, and azure-openai
as named LLM providers. Google Gemini is missing as a first-class option, even though
google-genai and litellm are already in requirements.txt.

Problem

Users wanting Gemini models (e.g. gemini-2.5-pro, gemini-1.5-flash) must manually
set up a LiteLLM proxy and use the generic openai-compatible provider — undocumented
and requiring extra infrastructure.

Proposed Solution

Add a gemini provider type:

• codewiki config set --provider gemini --api-key YOUR_GEMINI_KEY --main-model gemini-2.5-pro
• Direct integration via the existing google-genai SDK or litellm Gemini routing
• CLI examples in README

Files Affected

• codewiki/src/be/llm_services.py — add _call_llm_via_gemini(), route gemini provider
• codewiki/src/config.py — add gemini to the supported provider list
• README.md — add Gemini usage example

Acceptance Criteria

• --provider gemini accepted by codewiki config set
• Gemini models invokable without a LiteLLM proxy
• README updated with Gemini example