We require support for large xattr. 13500B is too low for some of our usages.
In this Jira, we are going to implement large XAttrs as a multi-row solution. That is, for each XAttr, we will split it into multiple rows with an index column indicating their appropriate order. Also, each row will include the total number of splits for that XAttr to simplify reading. We modified the hdfs_metadata_log table to include the inode name, parentId, and partitionId to avoid extra lookups in ePipe. Also, we ensure that setting metadata enabled on a directory that has existing XAttrs will result in XAttr logs added to the hdfs_metadata_log.