Hotfix for variant logic in the ApplicationExperiment

What does this MR do?

In reviewing the data we've tracked for the null_hypothesis experiment, we realized that there's a lack of expected memoization in gitlab-experiment experiments. This caused more events to be tracked for the candidate experience than we were expecting.

When we assign the control, there's a chance (because we don't memoize it) that on subsequent calls to resolve the variant again (for instance a call to track an event), we might incorrectly resolve and assign the candidate. This solves the issue as a hotfix, and will be tested and released in a new version of the gem.

Does this MR meet the acceptance criteria?

Conformity

• [-] Changelog entry
• Documentation (if required)
• Code review guidelines
• Merge request performance guidelines
• Style guides
• [-] Database guides
• Separation of EE specific content