Small improvements to modeling script

[HealthyPear]

This PR is the result of some improvement that came out as a consequence of the development of the integration tests for the modeling part.

• the split between train and test data has been simplified/improved by
  • using the correspondent scikit-learn function,
  • splitting on images instead of obs_id (more "democratic" choice)
  • shuffling the images against shower reuse in CORSIKA
• the I/O of the script has been improved,
  • input and output information can be overwritten by the CLI, which has lower priority over the config file
  • camera IDs can be read similarly with the following priorities (from higher to lower)
    • from the config file
    • from the training file
    • from the CLI

[codecov]

Codecov Report

Merging #114 (e87557c) into master (e708fe7) will decrease coverage by 0.61%.
The diff coverage is 9.43%.

@@            Coverage Diff             @@
##           master     #114      +/-   ##
==========================================
- Coverage   39.22%   38.61%   -0.62%     
==========================================
  Files          22       22              
  Lines        1912     1950      +38     
==========================================
+ Hits          750      753       +3     
- Misses       1162     1197      +35     

Impacted Files	Coverage Δ
protopipe/mva/train_model.py	15.68% <ø> (ø)
protopipe/scripts/build_model.py	11.29% <2.50%> (-3.93%)	⬇️
protopipe/pipeline/utils.py	50.90% <22.22%> (-1.66%)	⬇️
protopipe/mva/utils.py	12.31% <50.00%> (+0.97%)	⬆️

---

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 e708fe7...e87557c. Read the comment docs.