perf(core): Read the clock once per performance collection round (#5934)

* perf(core): Read the clock once per performance collection round

While a transaction is running, performance data is collected every
100ms. CompositeData.addDataAndCheckTimeout read the clock to check for
the 30 second collection timeout once per in-flight transaction,
allocating a SentryNanotimeDate each time, even though the enclosing
timer task had just read it for the sample it was distributing.

Pass that reading in, so a collection round reads the clock once instead
of once per transaction.

* changelog

Author: runningcode

CHANGELOG.md

@@ -16,6 +16,7 @@
 
 ### Performance
 
+- Read the clock once per performance collection round instead of once per in-flight transaction ([#5934](https://github.com/getsentry/sentry-java/pull/5934))
 - Reduce allocations while collecting cpu usage during transactions by reading the process cpu time via `Process.getElapsedCpuTime()` instead of parsing `/proc/self/stat` (33.6kB to 16 bytes per sample on a Pixel 3) ([#5926](https://github.com/getsentry/sentry-java/pull/5926))
 
 ### Dependencies

sentry/src/main/java/io/sentry/DefaultCompositePerformanceCollector.java

@@ -124,7 +124,7 @@ public void run() {
                 // Add the enriched tempData to all transactions/profiles/objects that collect data.
                 // Then Check if that object timed out.
                 for (CompositeData data : compositeDataMap.values()) {
-                  if (data.addDataAndCheckTimeout(tempData)) {
+                  if (data.addDataAndCheckTimeout(tempData, tempData.getNanoTimestamp())) {
                     // timed out
                     if (data.transaction != null) {
                       timedOutTransactions.add(data.transaction);
@@ -224,16 +224,19 @@ private CompositeData(final @Nullable ITransaction transaction) {
      * Adds the data to the internal list of PerformanceCollectionData. Then it checks if data
      * collection timed out (for transactions only).
      *
+     * @param nowNanos the timestamp of the current collection, passed in so a single clock reading
+     *     is shared by every transaction in this collection round.
      * @return true if data collection timed out (for transactions only).
      */
-    boolean addDataAndCheckTimeout(final @NotNull PerformanceCollectionData data) {
+    boolean addDataAndCheckTimeout(
+        final @NotNull PerformanceCollectionData data, final long nowNanos) {
       // stop() hands dataList out while this timer thread may still be writing to it, so consumers
       // synchronize on the list while iterating. We must hold the same monitor here.
       synchronized (dataList) {
         dataList.add(data);
       }
       return transaction != null
-          && options.getDateProvider().now().nanoTimestamp()
+          && nowNanos
               > startTimestamp
                   + TimeUnit.MILLISECONDS.toNanos(TRANSACTION_COLLECTION_TIMEOUT_MILLIS);
     }