fix(langchain): add support for LLMs without support for async callback handlers

[ajndkr]

Description

Fixes #164.

This PR adds new functionality to langchain adapter's streaming response: run mode.

Setting the run mode to "async" or "sync" defines how the langchain object is executed (via .acall or __call__ respectively)

Changelog:

• ✨ added run_mode attribute to StreamingResponse
• 📝 updated docs to create custom callback handlers