Since we're changing the pallet-mmr runtime & RPC APIs as part of #12339 it's a good opportunity to revisit some other decisions related to the API:

1. Whether we should have 2 separate methods generate_batch_proof() and generate_historical_batch_proof(). Since in the end generate_batch_proof() is a special case of generate_historical_batch_proof(). See the discussion here: pallet-mmr: generate historical proofs #12324 (comment)
2. If we should keep generate_proof(), verify_proof(), and verify_proof_stateless() since all these can be done via batching methods (e.g. generate_batch_proof(), etc)

cc: @acatangiu @Lederstrumpf @svyatonik