Unstable behavior in PredictRating() unit test due to constructor field shadowing

[ngtduc693]

Describe the bug

The unit test PredictRating_WithOtherUserHavingRatedTargetItem_ShouldCalculateSimilarityAndWeightedSum occasionally fails even though the logic and expected values are correct. The behavior is non-deterministic — it sometimes passes, sometimes fails.

To Reproduce

• Run the test suite multiple times (Run ALL unit tests, not only it)
• Observe that PredictRating_WithOtherUserHavingRatedTargetItem_ShouldCalculateSimilarityAndWeightedSum sometimes fails randomly

Expected behavior

• The field should consistently reference the injected dependency.
• The test should deterministically pass when the mock returns fixed values.

Actual behavior

• Test passes intermittently.
• When the mock reference is lost (null), similarity calculations fall back to undefined behavior.

Root cause

This test class used fixture-level fields (e.g., the ISimilarityCalculator mock, the CollaborativeFiltering instance, and the ratings data) that were re-initialized in [SetUp]. When the whole test suite runs (often in parallel), another test’s [SetUp] can overwrite those fields while this test is executing

=> Shared mutable test state + parallel execution = race condition.

[ngtduc693]

#534 I'm creating a PR to fix this issue, currently the PR is in draft state at the time of writing, once I test it more and it's stable I'll publish it