MAGE-974 InstantSearch extension points

[cammonro]

Summary

This PR aims to clean up the monolithic JS implementation with its heavy use of broadly scoped variables and closures and scattered implementation. We historically have had hooks to help customize the way the JavaScript works but it's very limited. Now it will be possible to write a mixin on any function to avoid overwriting the entire JS file.

• All logic has been split granularly into named, mixable functions
• Functions have been named in order to distinguish between InstantSearch widget objects, special facet config objects and builders
• XSS vulnerabilities were identified and corrected
• Behavior of conjunctive vs disjunctive facets has been clarified
• InstantSearch lib has been upgraded to v4.78
• Source maps have been added to InstantSearch to assist with debugging
• jQuery dependency has been significantly reduced (still not there - more work to do here but the goal is to remove altogether)
• Non relevant legacy code has been removed
• A discovered JS bug with empty hits has been resolved (this was impacting the execution chain and creating some weird behaviors with other widgets)
• Deprecation notices added - including UA sunsetted July 1, 2023
• Facet building operations have been refactored for clarity of intent
• A new front end hook called beforeFacetInitialization has been added which allows a user to extend the functionality by adding, removing or modifying "builder" functions which are used to define a facet config that will drive the rendering of facets

Result

Unit tests need to be added for this large revamp but manual testing has been performed including (but not limited to):

• General functionality
  
• XSS (and suggestions custom widget)
  
• nodeValue undefined bug fix
  
  
• Conjunctive facets (facet count change)
  
  
• Disjunctive facets (no facet count change)
  
  
• URL redirect on add to cart
  

[damcou]

Wonderful work ! 🥳 💪

I really like how the file looks now, this is a lot easier to understand, even for a non-js developer ;) .
Added a few comments but nothing blocking for you to go forward.

Out of curiosity, have you considered splitting the file into multiple ones since this is still a 1000+ lines one ? This isn't mandatory for sure, I guess this has its own pros.

[cammonro]

Thanks for the feedback! I'll pick up these observations on the next PR. 👍

Out of curiosity, have you considered splitting the file into multiple ones since this is still a 1000+ lines one ? This isn't mandatory for sure, I guess this has its own pros.

I did consider this! What's holding me back is that I would like to make these components available as UMD modules at some point for applications outside of Luma / RequireJS (like Hyvä etc.). Having too many dependencies can make that cumbersome unless we implement a build process for that (and we just ditched the build process necessary for algoliaBundle in the last release).

For now, I'm keeping the monolithic approach but just adding a bit of structure and scoping because before it was extremely brittle. Happy to revisit this in the future - but hopefully this is a step in the right direction.