Default nodetool tablestats to Human-Readable Output

Current Behavior

The current implementation of nodetool tablestats in Apache Cassandra outputs statistics in a format that is not immediately human-readable. This output primarily includes raw byte counts, which require additional calculation or conversion to be easily understood by users. This can be inefficient and time-consuming, especially for users who frequently monitor these statistics for performance tuning or maintenance purposes.

Proposed Change

We propose that nodetool tablestats should, by default, provide its output in a human-readable format. This change would involve converting byte counts into more understandable units (KiB, MiB, GiB). The tool could still retain the option to display raw data for those who need it, perhaps through a flag such as --no-human-readable or --raw.

Considerations

The change should maintain backward compatibility, ensuring that scripts or tools relying on the current output format can continue to function correctly.
We should provide adequate documentation and examples of both the new default output and how to access the raw data format, if needed.

Alignment

Discussion in the dev mailing list: https://lists.apache.org/thread/mlp715kxho5b6f1ql9omlzmmnh4qfby9 

Related work

Previous work in the series:

1. https://issues.apache.org/jira/browse/CASSANDRA-19015 
2. https://issues.apache.org/jira/browse/CASSANDRA-19104