[Search service] Integrate batched fetch (bfetch)

[lukasolson]

Since we are no longer batching search search requests using _msearch, we have started receiving reports that dashboards seem slower. We anticipate part of this is caused by hitting the browser connection limit (6 concurrent requests) when loading a large dashboard. To alleviate this problem, we'd like to integrate with bfetch, which batches multiple requests and streams responses as they become available.

One of the challenges we may encounter is in regards to cancelling in-progress requests in OSS. Since batching requests and streaming responses relies on keeping the connection open, we can't simply close the connection on a single request if the user navigates away or replaces the query. We also can't reliably send an additional request to cancel an individual query since in a multi-tenant scenario we can't reliably route to the server processing the request.

[elasticmachine]

Pinging @elastic/kibana-app-arch (Team:AppArch)

[lukasolson]

Closed by #84043.