[Profiling] Speed up processing of stacktraces

So far we kept track of stack frame and executable ids that are
associated to stack traces in a sorted collection. The rationale behind
this was that the mget should be faster. However, we have seen in
benchmarks that the contrary is the case: Not the mgets are the
bottleneck but rather insertion is so slow that it defers further
processing internally by several hundred milliseconds. When we change to
an unsorted (and pre-sized) collection we reduce median response time
from ~ 1200ms to 1000ms (or 17%).

x-pack/plugin/profiling/src/main/java/org/elasticsearch/xpack/profiling/TransportGetStackTracesAction.java

@@ -483,11 +483,8 @@ private class StackTraceHandler {
         private final GetStackTracesResponseBuilder responseBuilder;
         private final ActionListener<GetStackTracesResponse> submitListener;
         private final Map<String, StackTrace> stackTracePerId;
-        // sort items lexicographically to access Lucene's term dictionary more efficiently when issuing an mget request.
-        // The term dictionary is lexicographically sorted and using the same order reduces the number of page faults
-        // needed to load it.
-        private final Set<String> stackFrameIds = new ConcurrentSkipListSet<>();
-        private final Set<String> executableIds = new ConcurrentSkipListSet<>();
+        private final Set<String> stackFrameIds;
+        private final Set<String> executableIds;
         private final AtomicInteger totalFrames = new AtomicInteger();
         private final StopWatch watch = new StopWatch("retrieveStackTraces");
         private final StopWatch hostsWatch = new StopWatch("retrieveHostMetadata");
@@ -506,6 +503,9 @@ private StackTraceHandler(
             this.submitTask = submitTask;
             this.clusterState = clusterState;
             this.stackTracePerId = new ConcurrentHashMap<>(stackTraceCount);
+            // pre-size with a bit of headroom so the collection isn't resized too often
+            this.stackFrameIds = Collections.newSetFromMap(new ConcurrentHashMap<>(stackTraceCount * 5));
+            this.executableIds = Collections.newSetFromMap(new ConcurrentHashMap<>(stackTraceCount));
             this.expectedResponses = new AtomicInteger(expectedResponses);
             this.client = client;
             this.responseBuilder = responseBuilder;