Improve Ape-X DQN Implementation

[xiaomengy]

This PR introduced several improvements for Ape-X DQN listed below.

1. Use pre-computed q-value to compute priority. This is mentioned in the original Ape-X paper https://arxiv.org/abs/1803.00933
2. Add functional regularization from https://arxiv.org/abs/2106.02613
3. Add SpetralNorm from https://arxiv.org/abs/2105.05246
4. Modify default configs.