[Bug] Fix bug when dimension <= 56 in Faiss SQ.

[0ctopus13prime]

Description

The FaissSQDistanceComputer in faiss_sq_flat.h computes the 1-bit popcount dot product using only 8-byte (uint64_t) word loops: words = quantizedVectorBytes >> 3. When quantizedVectorBytes is not a multiple of 8, the trailing bytes are silently dropped. For example, dimension 56 produces 7 quantized bytes — words = 7 >> 3 = 0 — so the popcount loop never executes and the dot product is always zero. The HNSW graph is then built with no meaningful distance discrimination, resulting in ~16% recall. At dimension 57, Lucene's getDiscreteDimensions rounds up to 64 discretized dimensions (8 bytes), so words = 1 and the loop works correctly, giving ~88% recall. This creates a sharp recall cliff rather than a gradual degradation.

Fix:

Added a byte remainder loop in the IsBytesMultipleOf8=false branch of operator(), distances_batch_4, and symmetric_dis. After the 8-byte word loop, the remaining bytes are processed individually with __builtin_popcount. The IsBytesMultipleOf8=true path is unchanged — the remainder code only exists in the unaligned template instantiation, so there is zero performance impact for the common aligned case.

Related Issues

Resolves #[Issue number to be closed when this PR is merged]
N/A

Check List

• [O] New functionality includes testing.
• [O] New functionality has been documented.
• [O] API changes companion pull request created.
• [O] Commits are signed per the DCO using --signoff.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[naveentatikonda]

@0ctopus13prime pls fix the failing CI

Suite: Test class org.opensearch.knn.memoryoptsearch.faiss.FaissScalarQuantizedBulkSimdScorerTests
  2> REPRODUCE WITH: ./gradlew ':test' --tests 'org.opensearch.knn.memoryoptsearch.faiss.FaissScalarQuantizedBulkSimdScorerTests.testBBQCosineScoring' -Dtests.seed=66759F1A3FC1EA74 -Dtests.security.manager=false -Dtests.locale=ku -Dtests.timezone=Etc/GMT+12 -Druntime.java=21
  2> java.lang.AssertionError: Score mismatch at ord=224 for COSINE expected:<1.0> but was:<1.010622501373291>
        at __randomizedtesting.SeedInfo.seed([66759F1A3FC1EA74:517D2C11C2F232DE]:0)
        at org.junit.Assert.fail(Assert.java:89)
        at org.junit.Assert.failNotEquals(Assert.java:835)
        at org.junit.Assert.assertEquals(Assert.java:555)
        at org.opensearch.knn.memoryoptsearch.faiss.FaissScalarQuantizedBulkSimdScorerTests.doTest(FaissScalarQuantizedBulkSimdScorerTests.java:210)
        at org.opensearch.knn.memoryoptsearch.faiss.FaissScalarQuantizedBulkSimdScorerTests.testBBQCosineScoring(FaissScalarQuantizedBulkSimdScorerTests.java:69)

[0ctopus13prime]

Hey @naveentatikonda
Fixed it, thank you

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.02%. Comparing base (a1ee88c) to head (e743fb6).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##               main    #3229   +/-   ##
=========================================
  Coverage     83.02%   83.02%           
  Complexity     4160     4160           
=========================================
  Files           447      447           
  Lines         15313    15313           
  Branches       1964     1964           
=========================================
  Hits          12714    12714           
  Misses         1806     1806           
  Partials        793      793           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.