Services using gRPC will break when hitting the max gRPC message size limit

gRPC message default size limit (which I think was introduced in tonic 0.9) is set to 4MB.

With a lot of splits or indexes, a gRPC call to the metastore will return an error. For now, we saw a few examples happening on a quickwit cluster:

• a search request on tens thousand splits will hit the limit. The search request errors after calling list_splits (the merge planner too was concerned but we fixed the request to decrease the number of requested splits, see #3585)
• with an index with a lot of partitions, list_indexes_metadatas will error. This call is particularly critical.

Start by inspecting the metastore gRPC definitions and the implementations of list_splits and list_indexes_metadatas, then review how the tonic message-size limit is configured. Reproduce the failure with many splits or partitions and determine the intended handling for responses exceeding 4 MB. Done means both calls work for the reported large-index cases without hitting the limit.