HDDS-12079. Improve container merkle tree write handling

[errose28]

What changes were proposed in this pull request?

• Rename ContainerMerkleTree to ContainerMerkleTreeWriter to better reflect its purpose.
• Add data prefix to all checksum fields in the tree to differentiate them from other checksum types like metadata checksums that might be added later.
• Aggregate chunk checksums as they are added as leaves to the tree to reduce scanner memory footprint. We no longer need to hold every ChunkInfo object from the container in memory while the tree is being built.
• Use CRC32C to build the tree, regardless of the checksum used on the actual data.
  • CRC32C is a general improvement over CRC32, so we should use this to build the trees to prevent migrating later.
  • Per ChecksumByteBufferFactory#crc32CImpl, this will use a native CRC32C implementation for Java >= 9, otherwise fall back to a Java implementation.

What is the link to the Apache JIRA

HDDS-12079

How was this patch tested?

Existing tests must pass. No behavioral change is expected.

[aswinshakil]

Thanks for the patch @errose28. LGTM+1

[aswinshakil]

Thanks for the update @errose28. LGTM. Pending CI.