Implement alternative uncertainty propagation methods for full model uncertainty

[MoAly98]

Closes #221.
Blocked by #505.
This PR implements two alternative methods to propagate parameter uncertainties to a model, following the discussion in #221. The two implemented methods are:

• Error propagation by using the model's Jacobian (with finite-differences for derivatives) as implemented by iminuit
• Error estimation by bootstrapping the model predictions

Both methods can be accessed from the model_utils.yield_stdev function directly, or via the model_utils.prediction method. In both cases, the user simply specifies the name of the method that should be used for propagating the uncertainty. In case of bootstrapping, the user can also specify the seed for the bootstrap random number generator and the number of samples to be produced.

To-do:

• initial implementation of Jacobi method
• initial implementation of bootstrap method
• consistent variance array across all three methods of array propagation and harmonise transforming that to per-bin and per-channel quantities
• Modularise the code by splitting the methods into individual functions
• Wrap some common operations into modular methods
• Add tests
• Test closure in a model where closure is expected (model linear in all parameters, norm factors-only setup, all mods are linearly interpolated,..) -- See scikit-hep/pyhf/#2584

[github-actions]

Status: Blocked ❌

Issues blocking this PR:

• feat: Allowing the merging of samples for displaying model predictions without touching the PyHF model  #505 ❌

---

This comment was automatically written by the Blocking Issues bot, and this PR will be monitored for further progress.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (bb682b4) to head (6a8af67).

Additional details and impacted files

@@            Coverage Diff            @@
##            master      #535   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           22        22           
  Lines         2098      2159   +61     
  Branches       291       303   +12     
=========================================
+ Hits          2098      2159   +61     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.