Refactor Plotting functions

[abelaba]

This PR refactors the plotting functions by improving code modularity for the pairplot and marginal_pairplot functions and provides typed configuration using dataclasses. The changes include:

• Adds new dataclasses KdeDiagOptions, HistDiagOptions, ScatterDiagOptions passed to pairplot and marginal_pairplot functions, for customizing the diagonal plotting regions which are passed to the diag_kwargs argument.
• Adds new dataclasses KdeOffDiagOptions, HistOffDiagOptions, ScatterOffDiagOptions, ContourOffDiagOptions, PlotOffDiagKwargs passed to the pairplot function, for customizing the upper and lower plotting regions, which are passed to the upper_kwargs and lower_kwargs arguments.
• Adds a new dataclass FigOptions passed to pairplot and marginal_pairplot functions, for customizing the figure which is passed to the fig_kwargs argument.
• Moves helper functions to a separate file sbi/utils/plotting_helpers.py
• Modularizes repeated logic in pairplot and marginal_pairplot functions.
• Maintains backward compatibility with the previous dictionary-based configuration.
• Update the plotting tutorial to use the newly added dataclasses, docs/advanced_tutorials/17_plotting_functionality.ipynb.

This pull request builds upon the idea by @danielmk in PR #1529.

[codecov]

Codecov Report

❌ Patch coverage is 87.25869% with 33 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@ebb8c6f). Learn more about missing BASE report.
⚠️ Report is 8 commits behind head on main.

Files with missing lines	Patch %	Lines
sbi/analysis/plotting_classes.py	90.44%	13 Missing ⚠️
sbi/analysis/plot.py	82.25%	11 Missing ⚠️
sbi/utils/plotting_helpers.py	85.24%	9 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1631   +/-   ##
=======================================
  Coverage        ?   82.70%           
=======================================
  Files           ?      136           
  Lines           ?    11256           
  Branches        ?        0           
=======================================
  Hits            ?     9309           
  Misses          ?     1947           
  Partials        ?        0           

Flag	Coverage Δ
unittests	82.70% <87.25%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
sbi/utils/plotting_helpers.py	85.24% <85.24%> (ø)
sbi/analysis/plot.py	68.51% <82.25%> (ø)
sbi/analysis/plotting_classes.py	90.44% <90.44%> (ø)

[janosg]

This has almost no guarantee that valid kwargs come out.

asdict(replace(FigKwargs(), **fig_kwargs))

Would mean that the options from the user passed kwargs correspond to a field in FigKwargs and I think it would also run the any init or post init method that might do additional validation

[janosg]

even easier: asdict(FigKwargs(**fig_kwargs))

[janosg]

Here you choose to convert all possible ways of specifying kwargs to a dict; Why not convert everything to a Dataclass and then convert it to a dict in the last possible moment?

[janosg]

I would prefer the name DiagOptions; Kwargs sounds like a dictionary.

[janosg]

Are you sure all the options you have here need to be exposed in sbi? Matplotlib figures are mutable so many things can be changed after the figure is returned to the user. As long as data is plotted correctly and with the right color, most other things (e.g. titles) can be changed afterwards.