Change the approach for data structures: list/hash/sortedSet etc

Issue: Current benchmarking for data structures like hash, list, and sorted set is very poor, compared to string (as string DS, we have an **on-par performance with Redis**).

## Current Approach

For example, for the list data structure, when adding a new element to a `list` key, we have to do the below things:

- Client sends `lpush foo bar`
- Server gets byte array value from `foo` key from the DB
- Server encodes that byte array value to `LinkedList` java heap data structure
- Server adds new element `bar` to that `LinkedList`
- Server decodes that `LinkedList` object to byte array
- Server saves that new byte array into `foo` key in the DB

As you can see, decoding/encoding data structure from off-heap to on-heap is very costly, this leads to poor performance when the list is extensive, cost more heap space, and increase latency.