Enable prompt caching for direct OpenAI integration

[reynaldichernando]

A lot of our users upon landing on https://docs.puter.com/playground/, clicked run, but don't bother to wait for the result to show up

the main example we show is the gpt-5-nano model which does take a while to generate the text

it would be good if we have a way to have prompt caching so we could speed up this response

[ProgrammerIn-wonderland]

Will have to check if we support prompt cache for gpt nano (we definitely don't for responses yet) but this isn't a bad idea if it works