Batch generation of story cache may hit API rate limit

[AceZz]

especially for dalle

[Piwakk]

Then we will need to implement a "rate limiter"… for ourselves! But less complicated than what we have. In #152 everything is done concurrently (Promise.all), this might indeed be too ambitious. Even Firestore might not like this.

Better to batch everything:

• Record start time.
• Launch a batch of 5 (for instance)
• Wait for completion.
• Wait until now() - start_time > 1 minute (for instance).
• Repeat.

[AceZz]

let's do all the requests for a given form at once (this is only 27 requests). This should fit inside the open ai api rate limit (3rd line for us):

But let's wait for sure 1min between each formId when we generate in advance (next PR)