Review S3A documentation to make sure it is consistent with the current codebase

We need to review the hadoop-aws docs and make sure that it is consistent with the current codebase

• The performance.md doc is out of date w.r.t DDB on-demand and the threads, distcp (no -direct); maybe more.
• We should review the assumed role doc and combine the S3 and S3Guard permissions into one list
• the list of AWS regions is out of date (china 1 & 2, hong-kong, osaka)
• benchmark rename performance of large & huge files
• see if there is any current work on time-to-consistency for lists and updates

We also need to pull up the classpath problems "don't mix JARs unless you like stack traces" as the #1 entry on the list, as its the recurrent problem downstream, someone adding a new hadoop-aws JAR to a spark installation and then complaining about CNFEs