Change randomized prefix to contain only hex characters

[arhimondr]

Some storage systems prefer the prefix to be hexadecimal characters

No release notes entries required

[electrum]

Please add a comment explaining why this exact alphabet is important.

[arhimondr]

Certain storage systems prefer the prefix to be hexadecimal. Discussed offline.

[electrum]

Do we need to increase this now? This takes us from 2176782336 combinations to only 16777216, or 0.0077% of the previous, making the chance of collisions much higher. What are the implications of a collision?

[arhimondr]

S3 enforces request limits per prefix (https://aws.amazon.com/premiumsupport/knowledge-center/s3-request-limit-avoid-throttling/). The request limits are 3,500 PUT/COPY/POST/DELETE or 5,500 GET/HEAD requests. The space of 16777216 distinct prefixes should be sufficient.

[findepi]

Should these assumptions be persisted as a code comment, documenting why 6 is still sufficient prefix length?

[electrum]

Since we're enforcing hex, how about simplifying the generation function:

private static final int RANDOMIZED_PREFIX_BYTES = 3;

private static String generateRandomizedPrefix()
{
    byte[] bytes = new byte[RANDOMIZED_PREFIX_BYTES];
    ThreadLocalRandom.current().nextBytes(bytes);
    return BaseEncoding.base16().encode(bytes);
}

[arhimondr]

@electrum Having the custom function avoids unnecessary allocations and reduces uncertainty whether a lower case or upper case characters are used