refactor(accesscontrol): deterministic cache key hashing

[aruiz14]

Issue: rancher/rancher#47460

Problem

The current AccessFor implementation calculates the cache key for a given user based on the resource version of Roles/ClusterRoles associated via RoleBindings/ClusterRoleBindings. While this is correct, in the case of a cache miss that information that was previously used to calculate the hash is no used but instead retrieved again. Finally, the previously-calculated hash is used as key to store the generated AccessSet into the cache.

This approach Introduces the chance for a race condition, which this PR is meant to fix by using the same data used to calculate the hash to later generate the AccessSet in case of a cache miss.

Summary

• Small refactoring of some internal functions to allow reusing them
• Add new types for encapsulating all needed data
• Make accessStore use the new types and method
• Cleanup unused code and adapt tests to new signatures
• Add new comments to structs and methods
• Refactor PolicyRules' indexer to make it easier to read and more testable

(please take a look at the commits view for more details)

[ericpromislow]

Looks fine, apart from a trivial typo

[ericpromislow]

thx

[bigkevmcd]

I've gone over this in fine detail, and it looks fine.

It improves the test coverage a fair bit, but there's still a lot of untested code that we can address when we next touch these files.

[aruiz14]

@ericpromislow please take another look, I've addressed Kevin's feedback but it didn't affect core of this PR.

[ericpromislow]

Works and looks ok