fix: make the redis cluster e2e test target run against the test stack

[alexluong]

Makes make up/test/rediscluster + make test/e2e/rediscluster run the Redis Cluster e2e suite again. It had drifted and either failed to start or ran no tests.

1. test/e2e/rediscluster ran -run TestRedisClusterBasicSuite, which no longer exists, so it matched nothing. Now -run TestE2E_Compat_RedisCluster.
2. The test-runner pointed at hostnames that don't exist on the test network (postgres, clickhouse, mock, …). Now it uses the make up/test containers (test-postgres, test-clickhouse, test-mock) and adds TEST_NATS_URL, which e2e needs since NATS became the default internal MQ. The RabbitMQ/LocalStack/GCP/Azure entries pointed at nothing and are gone.
3. TESTCOMPAT=1 is set on the test-runner, since the suite skips without it.
4. grokzen/redis-cluster:7.2.4 sets a locale it doesn't ship and Redis exits on start, so the cluster never came up on Linux. LC_ALL=C fixes it.
5. contributing/test.md documented a TESTREDISCLUSTER variable that nothing reads; removed.

With this, the suite runs end to end. 4 subtests fail (TestRetry_FailedDeliveryAutoRetries, TestRetry_ManualRetryCreatesNewAttempt, TestRetry_DuplicateManualRetryExecutesBoth, TestOpEvents_ExhaustedRetries) on a real bug: the retry queue's keys span two cluster slots. The fix is #1139, stacked on this one.

To run: make up/test && make up/test/rediscluster && make test/e2e/rediscluster.

🤖 Generated with Claude Code