Connector should validate 'is_write_index' flag on aliases during write operations

[jbaiera]

ES-Hadoop has been historically very hands off when it comes to aliases. Recently added in 6.X is the ability to specify an index within an alias to be used for accepting write operations.

We should check the write resource when the job starts up to detect if an alias is being used for the write operation, and validate that either there is only one index within it, or if multiple indices are given for an alias, that one is marked as the write index. This will allow a job to fail faster than if an alias cannot be written to at all. Additionally, we should not attempt to pin to any nodes that are hosting the shards for those writes, since the fact that an alias is used for writing means that the shards that we are writing to might change from under us at any time, making the original node pinning not as useful.