controller: use PageserverUtilization for scheduling (#8711)

## Problem

Previously, the controller only used the shard counts for scheduling.
This works well when hosting only many-sharded tenants, but works much
less well when hosting single-sharded tenants that have a greater
deviation in size-per-shard.

Closes: https://github.com/neondatabase/neon/issues/7798

## Summary of changes

- Instead of UtilizationScore, carry the full PageserverUtilization
through into the Scheduler.
- Use the PageserverUtilization::score() instead of shard count when
ordering nodes in scheduling.

Q: Why did test_sharding_split_smoke need updating in this PR?
A: There's an interesting side effect during shard splits: because we do
not decrement the shard count in the utilization when we de-schedule the
shards from before the split, the controller will now prefer to pick
_different_ nodes for shards compared with which ones held secondaries
before the split. We could use our knowledge of splitting to fix up the
utilizations more actively in this situation, but I'm leaning toward
leaving the code simpler, as in practical systems the impact of one
shard on the utilization of a node should be fairly low (single digit
%).

storage_controller/src/node.rs

@@ -92,15 +92,15 @@ impl Node {
         }
     }
 
-    pub(crate) fn get_availability(&self) -> NodeAvailability {
-        self.availability
+    pub(crate) fn get_availability(&self) -> &NodeAvailability {
+        &self.availability
     }
 
     pub(crate) fn set_availability(&mut self, availability: NodeAvailability) {
         use AvailabilityTransition::*;
         use NodeAvailability::WarmingUp;
 
-        match self.get_availability_transition(availability) {
+        match self.get_availability_transition(&availability) {
             ToActive => {
                 // Give the node a new cancellation token, effectively resetting it to un-cancelled.  Any
                 // users of previously-cloned copies of the node will still see the old cancellation
@@ -115,8 +115,8 @@ impl Node {
             Unchanged | ToWarmingUpFromOffline => {}
         }
 
-        if let (WarmingUp(crnt), WarmingUp(proposed)) = (self.availability, availability) {
-            self.availability = WarmingUp(std::cmp::max(crnt, proposed));
+        if let (WarmingUp(crnt), WarmingUp(proposed)) = (&self.availability, &availability) {
+            self.availability = WarmingUp(std::cmp::max(*crnt, *proposed));
         } else {
             self.availability = availability;
         }
@@ -126,12 +126,12 @@ impl Node {
     /// into a description of the transition.
     pub(crate) fn get_availability_transition(
         &self,
-        availability: NodeAvailability,
+        availability: &NodeAvailability,
     ) -> AvailabilityTransition {
         use AvailabilityTransition::*;
         use NodeAvailability::*;
 
-        match (self.availability, availability) {
+        match (&self.availability, availability) {
             (Offline, Active(_)) => ToActive,
             (Active(_), Offline) => ToOffline,
             (Active(_), WarmingUp(_)) => ToWarmingUpFromActive,
@@ -153,15 +153,15 @@ impl Node {
 
     /// Is this node elegible to have work scheduled onto it?
     pub(crate) fn may_schedule(&self) -> MaySchedule {
-        let score = match self.availability {
-            NodeAvailability::Active(score) => score,
+        let utilization = match &self.availability {
+            NodeAvailability::Active(u) => u.clone(),
             NodeAvailability::Offline | NodeAvailability::WarmingUp(_) => return MaySchedule::No,
         };
 
         match self.scheduling {
-            NodeSchedulingPolicy::Active => MaySchedule::Yes(score),
+            NodeSchedulingPolicy::Active => MaySchedule::Yes(utilization),
             NodeSchedulingPolicy::Draining => MaySchedule::No,
-            NodeSchedulingPolicy::Filling => MaySchedule::Yes(score),
+            NodeSchedulingPolicy::Filling => MaySchedule::Yes(utilization),
             NodeSchedulingPolicy::Pause => MaySchedule::No,
             NodeSchedulingPolicy::PauseForRestart => MaySchedule::No,
         }
@@ -285,7 +285,7 @@ impl Node {
     pub(crate) fn describe(&self) -> NodeDescribeResponse {
         NodeDescribeResponse {
             id: self.id,
-            availability: self.availability.into(),
+            availability: self.availability.clone().into(),
             scheduling: self.scheduling,
             listen_http_addr: self.listen_http_addr.clone(),
             listen_http_port: self.listen_http_port,