[ML] Use optimal test parallelism based on CPU count

Each test suite spawns ctest --parallel <all_cpus> internally, so
running all suites concurrently causes resource contention. Limit
the number of concurrent suites to ceil(nproc/3) on Unix and 2 on
Windows, matching empirically determined optimal values from
cross-platform benchmarking (see PR elastic#2900).

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: edsavage

build.gradle

@@ -110,6 +110,11 @@ if (isWindows && msystem != 'MINGW64') {
 }
 
 project.ext.numCpus = Runtime.runtime.availableProcessors()
+// Optimal number of test suites to run concurrently via cmake --build -j.
+// Each suite spawns ctest --parallel <all_cpus> internally, so running too
+// many suites simultaneously causes resource contention. These values were
+// determined empirically (see PR #2900).
+project.ext.testParallel = isWindows ? 2 : Math.max(1, (int) Math.ceil(numCpus / 3.0))
 project.ext.makeEnvironment = [ 'CPP_CROSS_COMPILE': cppCrossCompile,
                                 'VERSION_QUALIFIER': versionQualifier,
                                 'SNAPSHOT': (isSnapshot ? 'yes' : 'no'),

dev-tools/docker/docker_entrypoint.sh

@@ -110,9 +110,3 @@ elif [ "x$1" = "x--test" ] ; then
     cmake --build cmake-build-docker ${CMAKE_VERBOSE} -j ${TEST_PARALLEL} -t test_all_parallel || echo failed > build/test_status.txt
 fi
 
-# Print sccache stats if it was used
-if [ -n "${SCCACHE_PATH:-}" ]; then
-  "$SCCACHE_PATH" --show-stats || true
-  "$SCCACHE_PATH" --stop-server || true
-fi
-