cluster: ignore _last_replicate value in destructor

Last replicate future is an internal barrier for sync operations. Ignore
its value if we're shutting down to prevent seastar logging warnings for
unhandled exception.

Couldn't come up with a test for this.

src/v/cluster/archival_metadata_stm.cc

@@ -599,6 +599,15 @@ archival_metadata_stm::archival_metadata_stm(
   , _cloud_storage_api(remote)
   , _feature_table(ft) {}
 
+archival_metadata_stm::~archival_metadata_stm() {
+    // Last replicate future is an internal barrier for sync operations. Ignore
+    // its value if we're shutting down to prevent seastar logging warnings for
+    // unhandled exception.
+    if (_last_replicate.has_value()) {
+        _last_replicate->ignore_ready_future();
+    }
+}
+
 ss::future<std::error_code> archival_metadata_stm::truncate(
   model::offset start_rp_offset,
   ss::lowres_clock::time_point deadline,

src/v/cluster/archival_metadata_stm.h

@@ -111,6 +111,7 @@ class archival_metadata_stm final : public raft::persisted_stm<> {
       features::feature_table&,
       ss::logger& logger,
       ss::shared_ptr<util::mem_tracker> partition_mem_tracker = nullptr);
+    ~archival_metadata_stm() override;
 
     /// Add segments to the raft log, replicate them and
     /// wait until it is applied to the STM.