Consider a better default implementation of DefaultPartitionSelector that will work with UUIDs

Describe the issue
The hash algorithm used in org.springframework.cloud.stream.binder.PartitionHandler.DefaultPartitionSelector is not suited for UUIDs combined with a small number of partitions divisible by 8.

We are using UUIDs as the key for messages. What we found is that we got a very uneven distribution across our 24 partitions.

According to Grok:
If the lower bits of the hash code are not uniformly random (due to the XOR operation or UUID structure), then modulus operations with powers of 2 (like 8) will over-represent some buckets and under-represent others.

We could confirm that this issue happens with a small number of partitions where the number of partitions is divisible by 8.

To Reproduce
See following tests with numberOfPartitions or 8, 16, 24.

   @Test
    void testDefaultPartitionSelector() {
        final int numberOfPartitions = 8;
        for (int i = 0; i< 20; i++) {
            String key = UUID.randomUUID().toString();
            System.out.println("Key: " + key + " " + "Partition: " +Math.abs(key.hashCode() % numberOfPartitions));
        }
    }

    @Test
    void testDefaultPartitionSelectorWithSpecificKeys() {
        final int numberOfPartitions = 8;
        String[] keys = {
                "f91f585a-2c4d-470c-a458566bcefea2d7",
                "79f29968-3c16-4139-b660d4ac39b75b0f",
                "5900c406-6b64-49dd-a88d7b4f738e1cc4",
                "2e79f680-adfc-480b-a0d8d3bb3b767020",
                "cf8159d1-2318-4fc8-9f7bbd05774c1b21",
                "7ac3ecbc-6516-4a92-9c3616a36c7fc11b"
        };
        for (String key : keys) {
            System.out.println("Key: " + key + " " + "Partition: " +Math.abs(key.hashCode() % numberOfPartitions));
        }
    }

Version of the framework
2023.0.4

Expected behavior
Even distribution of messages across partitions

Screenshots

Additional context
Kafka uses MurmurHash to determine partitions and that appears to work much better. Using the following override works better than the default implementation.

    @Bean
    public PartitionSelectorStrategy murmur2PartitionSelectorStrategy() {
        return (key, partitionCount) -> key != null ? Utils.toPositive(Utils.murmur2(key.toString().getBytes(StandardCharsets.UTF_8))) : 0;
    }

Start with org.springframework.cloud.stream.binder.PartitionHandler.DefaultPartitionSelector and reproduce the uneven distribution using the two test cases in the issue with 8, 16, and 24 partitions. Compare the current behavior with the Kafka MurmurHash-based approach described in the issue. Done means UUID keys are distributed more evenly across partition counts divisible by 8, with regression coverage for the reported cases.