Propagate standby apply LSN to pageserver to hold off GC.

To avoid pageserver gc'ing data needed by standby, propagate standby apply LSN
through standby -> safekeeper -> broker -> pageserver flow and hold off GC for
it. Iteration of GC resets the value to remove the horizon when standby goes
away -- pushes are assumed to happen at least once between gc iterations. As a
safety guard max allowed lag compared to normal GC horizon is hardcoded as 10GB.
Add test for the feature.

Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>

storage_broker/benches/rps.rs

@@ -147,6 +147,7 @@ async fn publish(client: Option<BrokerClientChannel>, n_keys: u64) {
                 http_connstr: "zenith-1-sk-1.local:7677".to_owned(),
                 local_start_lsn: 0,
                 availability_zone: None,
+                standby_horizon: 0,
             };
             counter += 1;
             yield info;

storage_broker/proto/broker.proto

@@ -42,6 +42,7 @@ message SafekeeperTimelineInfo {
     uint64 remote_consistent_lsn = 7;
     uint64 peer_horizon_lsn = 8;
     uint64 local_start_lsn = 9;
+    uint64 standby_horizon = 14;
     // A connection string to use for WAL receiving.
     string safekeeper_connstr = 10;
     // HTTP endpoint connection string
@@ -105,4 +106,6 @@ message SafekeeperDiscoveryResponse {
     string safekeeper_connstr = 4;
     // Availability zone of a safekeeper.
     optional string availability_zone = 5;
+    // Replica apply LSN
+    uint64 standby_horizon = 6;
 }

storage_broker/src/bin/storage_broker.rs

@@ -736,6 +736,7 @@ mod tests {
             http_connstr: "neon-1-sk-1.local:7677".to_owned(),
             local_start_lsn: 0,
             availability_zone: None,
+            standby_horizon: 0,
         })
     }