[Monitoring] Only perform a single date_histogram in get_nodes

[chrisronline]

This comment discovered that we're doing more date_histogram aggregations than necessary in our get_nodes call.

We need to fix this, and also adjust the relevant server code to convert the updated aggregation response to the same as before, so all of the consumers continue to work properly.

[elasticmachine]

Pinging @elastic/stack-monitoring