Should cudf hash decimal values the same regardless of scale?

Currently, cudf hashes decimals the same way regardless of their scale. To me, this is a violation of "different values hash differently". The scale is a piece of column metadata but I think it's relevant to the hash computation because hashes are typically used as a proxy for equivalent representations. I think this could be considered a bug. My proposal would be to combine the column scale into the hash of the value.

_Originally posted by @bdice in https://github.com/rapidsai/cudf/pull/13612#discussion_r1264530164_