Feature description

Enhance the _call_pipeline method to support batched evaluation. Instead of processing the whole dataset at once, the method should allow to group dataset items into batches and process them sequentially.

Motivation

APIs like e.g. Azure OpenAI imposes rate limits on the number of requests a user or client can make within a specified period. Currently, _call_pipeline processes the whole dataset at once, which can quickly exceed these limits.

Additional context

No response