Pull HSI_Event class into its own submodule

[willGraham01]

This pull request moves the HSI_Event class (and related supporting classes) out of the healthsystem module and into its own submodule (hsi_event). This allows us to:

• Add typehints to the HSI_Event class.
• Perform some simplifications to the internal class logic (notably not storing objects that are not referenced, or can be obtained through properties rather than storing copies of the original object).
• These changes are largely intended to help make implementing changes for A lot of time spent in do_at_generic_first_appt_non_emergency due to repeated dataframe access #1237 simpler and slightly less noisy.

There are no functionality changes that are introduced by this PR.

Key changes

• The HSI_Event class and related container classes have been moved into the src/tlo/methods/hsi_event.py file, to make them accessible without needing to import the entire healthcaresystem.py module.
  • This also lets us avoid circular imports that previously prevented type-hinting.
• Internal logic of the HSI_Event class has been simplified slightly in places:
  • Class attributes are given typehints.
  • Avoids storing the parent Module, Simulation, and Population.
    • We now just store the Module.
    • The simulation can be accessed through a self.sim property to avoid storing it as a class attribute.
    • The healthcare system can also be accessed through a shortcut self.healthsystem property.
    • The target_is_alive property is used in place of tracking the population dataframe as a class attribute. We were only looking up one value from the dataframe, and otherwise not using it at all.
  • Avoids dynamically creating the return_item_codes_in_dict function every time the get_consumables method is run.
    • return_item_codes_in_dict is now a stand-alone method.
  • adjust_facility_level_to_merge_1b_and_2 has been made into a class method since it is not used outside of the HSI_Event class.
  • Typehints have been added to member functions and attributes where possible.

[willGraham01]

/run profiling

This PR should be performance agnostic compared to master - it is just a refactoring.

[ucl-comment-bot]

Profiled run of the model succeeded ✅

🆔 22554782937
⏲️ 591 minutes
#️⃣ 1972781

[matt-graham]

Thanks @willGraham01 these all looks like sensible changes to make. I have made a few small suggestions though none are vital.

Just so its recorded somewhere, Will and I discussed in person the changes to the imports of HSI_Event in other modules, which are not strictly necessary as HSI_Event is imported from the new tlo.methods.hsi_event module in tlo.methods.healthsystem and so could technically still be imported from there, which would keep the changes in this PR confined to the healthsystem and new hsi_event modules. Will pointed out changing the imports helps avoid circular imports between disease modules and healthsystem and should simplify importing HSI_Event to use just for type hinting purposes without needing any if TYPE_CHECKING blocks, which I felt was a good enough justification to outweigh the increased probability of needing to deal with some merge conflicts in disease modules on other branches when this is merged in (though as this only touches imports shouldn't be too much of a problem).

[willGraham01]

Have adapted the type-hinting to use from future and moved the stray method into the HSI_Event class to hide it. Re-requesting review in case there's anything else, otherwise will merge when (/if!) tests pass

[matt-graham]

Thanks for doing the updates