fix: lower NFR throughput CI thresholds from 500 to 300 ops/sec (#1550)

GitHub Actions runners consistently achieve 390-470 ops/sec on these
benchmarks, causing flaky failures at the 500 threshold. Lower to 300
to provide headroom while still catching real performance regressions.
Production targets remain unchanged (10000 and 1000 ops/sec).

Co-authored-by: Ben Coombs <bjcoombs@users.noreply.github.com>

Author: bjcoombs

services/internal-account/benchmarks/load_test.go

@@ -275,9 +275,10 @@ func TestThroughputUnderLoad(t *testing.T) {
 	}
 
 	// Advisory target: >1000 ops/sec
-	// CI environments may be slower due to shared resources
-	if opsPerSecond < 500 {
-		t.Errorf("Throughput below CI threshold: %.0f (CI threshold: >500, production target: >1000)", opsPerSecond)
+	// CI environments may be slower due to shared resources.
+	// CI runners consistently achieve 390-470 ops/sec; 300 provides headroom.
+	if opsPerSecond < 300 {
+		t.Errorf("Throughput below CI threshold: %.0f (CI threshold: >300, production target: >1000)", opsPerSecond)
 	} else if opsPerSecond < 1000 {
 		t.Logf("Note: Throughput %.0f below production target (>1000) but passes CI threshold", opsPerSecond)
 	}

services/internal-account/benchmarks/nfr_validation_test.go

@@ -430,8 +430,9 @@ func TestNFR_SustainedThroughput(t *testing.T) {
 
 	// Advisory throughput target - testcontainer environments have significant overhead.
 	// Production with connection pooling achieves much higher throughput.
+	// CI runners consistently achieve 390-470 ops/sec; 300 provides headroom.
 	targetThroughput := 10000.0
-	ciThreshold := 500.0 // Relaxed for CI testcontainer overhead
+	ciThreshold := 300.0 // Relaxed for CI testcontainer + shared runner overhead
 
 	if opsPerSecond < ciThreshold {
 		t.Errorf("Throughput below CI threshold: %.0f (CI threshold: >%.0f, production target: >%.0f)", opsPerSecond, ciThreshold, targetThroughput)