#9321 introduced batching of get page requests.
One get vectored request handles the entire batching.

While batching we call PageServerHandler::wait_or_get_last_lsn to determine the effective LSN
of a request. This pushes the wait further up the stack, which isn't a problem for single get page requests,
but delays the entire batch if there's multiple of them.

We can do better: stop batching if we have to wait and process what we've collected. We can achieve this
by splitting PageServerHandler::wait_or_get_last_lsn in two separate methods: one to resolve the LSN
and a second one to wait. Once we have that update the batching logic.