Pipeline for no variability

[Flaminietta]

I'd like to open an intermediary PR.

In this PR all the normalization functions (the ones present in the R-code) are implemented and tested. For now, I am considering only an MCDA run with no variability analysis (no Monte Carlo sampling).

The next steps will then be:

• add all aggregation functions - on the same branch
• implement pipeline with MC sampling - new branch
• add more norm/agg functions - maybe later?

Please, have a look at the logic in particular, by testing I've seen that it is simple to make distraction errors in the implementation of the functions.

[Flaminietta]

@mspada What I still did not do for all the norm Python functions is to test them against the R-functions, something I can easily do next. I'll report on these tests soon.

[Flaminietta]

All normalization functions have been tested against the ones implemented in R. minmax reverse is now correct. All norm functions produce the same results in both implementation.

[mspada]

Thanks Flaminia, all fine to me

[Flaminietta]

@mspada @kapil-agnihotri

The PR is now ready for your final review. It adds:

• MCDA without variability (you can run it with python3 -m mcda.mcda_run -c configuration_without_uncertainty.json)
  • note: change the output path
  • it will save the scores, the norm. scores, the ranks, a copy of the configuration file, and a plot of the scores in a new directory
• unit tests for all the major functionalities (run them all with: python -m pytest -s tests/unit_tests/)
• note: I have noticed that if you open the output file ranks.csv with e.g. Numbers sometimes the values are incorrect (the decimal point disappears; this, however, does not happen in e.g. Excel and it seems not to be a problem within Python)

I'll squash the commits during the merge.