Keeping Your Distance: Solving Sparse Reward Tasks Using Self-Balancing Shaped Rewards