Don't retry flakes in integration tests

Flaky behavior can hurt real users, and ignoring it to appease CI tests
only hides the problem, and can make CI tests take longer.

If this is too noisy, we can either spend effort fixing flakes, and/or
make tests less strict, or re-enable flaky retries.

Makefile

@@ -33,7 +33,7 @@ TEKTON_VERSION ?= v0.20.1
 SDK_VERSION ?= v0.17.0
 
 # E2E test flags
-TEST_E2E_FLAGS ?= -failFast -flakeAttempts=2 -p -randomizeAllSpecs -slowSpecThreshold=300 -timeout=30m -progress -stream -trace -v
+TEST_E2E_FLAGS ?= -failFast -p -randomizeAllSpecs -slowSpecThreshold=300 -timeout=30m -progress -stream -trace -v
 
 # E2E test operator behavior, can be start_local or managed_outside
 TEST_E2E_OPERATOR ?= start_local
@@ -219,7 +219,6 @@ test-integration: install-apis ginkgo
 		-randomizeAllSpecs \
 		-randomizeSuites \
 		-failOnPending \
-		-flakeAttempts=2 \
 		-slowSpecThreshold=240 \
 		-trace \
 		test/integration/...