fix: increase default LLM timeout to 300s for local models

[NotYuSheng]

Summary

• The default LLM_TIMEOUT of 60s was too short for local LLMs (e.g. LM Studio with Qwen2.5-14B), causing story generation to fail with "LLM server is not responding" even when the model was actively processing
• Added LLM_TIMEOUT=300 to .env.example and docker-compose.yml, raising the default to 5 minutes
• Users can override via the LLM_TIMEOUT env var if their hardware needs more time

Test plan

• Set LLM_TIMEOUT=300 in .env, restart backend with docker compose up -d backend
• Trigger story generation on a PCAP file with a local LM Studio model
• Confirm story generates successfully without timeout error

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request introduces a configurable LLM_TIMEOUT environment variable in .env.example and docker-compose.yml, defaulting to 300 seconds to accommodate slower local models. Review feedback points out that the implementation is incomplete because the internal application defaults in application.yml and LlmConfig.java are still set to 60 seconds, which may lead to inconsistent behavior. Additionally, it is suggested to update the documentation in .env.example to clarify that 300 seconds is a recommended value rather than the system-wide fallback.