Gsmap

- add run_visualize subcommand to visualize results.
- correct some typos.

- add a function, which could use enhancer annotation.
- avoid overflow of float16

- optimize the latent to gene mem usage
- add --no_expression_fraction option

Improve:
1. Reduce the Spatial LDSC Step RAM Usage
2. Speed up the Latent to Gene Step
3. Use Sparse Matrix For generate LD score
4. Some Bug fixed