Add parallel processing to map step of map reduce summarisation

[andy-symonds]

Context

Before this change, the map step of map reduce summarisation, used for large document summarisation, was not happing in parallel. The code needed to be updated to allow the parallele summarisation of large documents into smaller summaries using LCEL batch.

batch only takes a list as input, so I needed to split the question prompt into map_question_prompt, which is populated with the question and chat_history before being passed to the map_operation and then the map_document_prompt which is populated when running batch.

Changes proposed in this pull request

• Updated build_map_reduce_summary_chain to allow for batch processing
• Made Chunk model iterable
• Renamed max_tokens in env var AI Settings to summarisation_chunk_max_tokens to better describe what it does. max_tokens is already an overloaded term.

Guidance to review

Check you are happy with how I have structured build_map_reduce_summary_chain.

Relevant links

Things to check

• I have added any new ENV vars in all deployed environments
• I have tested any code added or changed
• I have run integration tests

[jamesrichards4]

Looks good. Probably worth making max_concurrency a setting so we can switch it out if we get issues though