List the major blocking calls in current implementation and alternatives/workarounds based on async

• Generic http requests: use aiohttp because it's fast and lightweight
• Azure table operations: directly implement asyncio, sample code
• Azure Speech SDK: It appears we're already using some kind of non-blocking API called speak_ssml_async
• OpenAI SDK: there seems to be a PR that enables async calls, or we can just convert everything to aiohttp requests, I guess, since we're planning to use it anyway...
• time.sleep(): ah, a classic. asyncio.sleep() will do.