chore(internal): set up TestContainers for running mock server

[jdubois]

• This removes the need to have NPM installed
• This should be faster as there is no "npm install"
• This removes the risk of having a mock server that keeps running in the background after the tests have benn run

Fix #54

[TomerAberbach]

I am realizing that we use our own custom fork of Prism...

openai-java/scripts/mock

Line 24 in 908c42c

npm exec --package=@stainless-api/[email protected] -- prism mock "$URL" &> .prism.log &

Is it true that we'd need to publish our own Docker image of that to use it here?

[jdubois]

Oh why are you doing this @TomerAberbach ? That's a bit surprising, especially as I have all the tests working fine with the "standard" Prism. You can try this PR, everything is OK!
Then if that's what you want, I can still make this work in TestContainers, even without a customer image: it's just a matter of running NPM inside Docker. It'll just be a bit slower (time to pull everything), and not as much reproductible (if there's a network issue when pulling an NPM package)

[TomerAberbach]

I think the reason is because Prism has a bunch of bugs and weird behaviors that we've had to fix over time.

It's quite possible that none of these issues currently affect OpenAI, but they affect other Stainless customers, and so they may affect OpenAI in the future as the spec is updated.

If it's not too much trouble, making it work with the fork would be amazing 🙏