`PromptNode` doesn't support isolated node evaluation

[Timoeller]

Describe the bug When running pipeline.eval_batch with a GenerativeQAPipeline the isolated node eval options seems to be ignored.

Additional context

1. When running the same pipeline without batching it adds the isolated node eval. Only in eval_batch it dissmisses the parameter.
2. Running eval_batch with an ExtractiveQAPipeline does not ignore the isolated eval param.

To Reproduce please take eval labels from the eval tutorial

    retriever = BM25Retriever(
        document_store=document_store
    )
    generator = Seq2SeqGenerator(model_name_or_path="vblagoje/bart_lfqa")
    pipeline = GenerativeQAPipeline(generator=generator, retriever=retriever)
    eval_result = pipeline.eval_batch(labels=eval_labels[:5],
                                      params={"Retriever": {"top_k": 10}, "Generator": {"top_k": 2}},
                                      sas_model_name_or_path="cross-encoder/stsb-roberta-large",
                                      add_isolated_node_eval=True)

[bogdankostic]

#5291 adds support for isolated node evaluation in BaseGenerator.

Given that we removed Seq2SeqGenerator and RAGenerator and recommend to use PromptNode instead, I investigated the support for isolated node evaluation in PromptNode. However, I found that we currently do not support it at all. Let's keep this issue open to think about if and how we want to support isolated node evaluation in PromptNode.

[masci]

Closing as won't fix