Add batching to redis_hash output to improve performance

I am starting to experience performance problems when using the redis_hash output at a high number operations/s. The current implementation of the redis_hash output does not utilize the pipelining feature of redis-go. Adding this, similar to other redis outputs might be a good idea.

I have little experience with go, but I am willing to attempt this, staying close to the original implementation and borrowing parts of the pipelining implementation of the redis-list output.

Start by locating the redis_hash output, then compare it with the redis-list output's pipelining implementation and inspect how redis-go supports pipelining. Add batching to redis_hash while staying close to its current implementation, and verify that the output uses pipelining as intended.