Add initial load tests

[jackapbutler]

• closes #1622

I've used locust to write a basic load test which will hit two endpoints sequentially that mimic the text-client in the inference server.

1. /chat to start a new conversation with chat_id
2. /chat/{chat_id}/message to send a message to the Assistant

An isolated load test user workflow is summarised by first spawning $X$ users every $T$ seconds to a maximum amount of $N$ concurrent users

1. A user starts a conversation with the Assistant
2. Then they enter a conversation loop
   1. Send a chat message to the Assistant
   2. Wait until Assistant responds
   3. Wait $S$ further seconds
   4. Repeat

Initial results showed the Bot failing (potentially due to a race condition) and I'm currently investigating the source of that error. The main error happens inside of add_prompter_message() and is is caused by the chat.pending_message_request status being non None giving rise to a HTTPError of “Already pending”.