Enable uncertainty quantification using quantile loss

[kailingding]

Summary

• torchts
  • added quantile_loss
  • added criterion_args parameter in TimeSeriesModel class in order to enable users to pass in customized quantile
  • added unit test for loss.py
• examples
  • added an example notebook that shows how to use torchts to make prediction intervals

Reviewer Request

@klane

Paper

https://dl.acm.org/doi/pdf/10.1145/3447548.3467325

[yuqirose]

@DongxiaW, pls review this commit.

[codecov]

Codecov Report

Merging #168 (53c7081) into main (27071ae) will increase coverage by 3.61%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #168      +/-   ##
==========================================
+ Coverage   18.52%   22.13%   +3.61%     
==========================================
  Files           7        7              
  Lines         394      402       +8     
==========================================
+ Hits           73       89      +16     
+ Misses        321      313       -8     

Impacted Files	Coverage Δ
torchts/nn/loss.py	100.00% <100.00%> (+100.00%)	⬆️
torchts/nn/model.py	79.66% <100.00%> (+1.08%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 27071ae...53c7081. Read the comment docs.