Wrap dropped-FP scalar-quantized float values so scoring stays quantized

[Pulkitg64]

Description

When an index has had its raw full-precision float vectors dropped (the read-only "drop-FP" / searcher-node optimization, where the flat raw-vector file is empty and only the scalar-quantized bytes remain), Lucene104ScalarQuantizedVectorsReader#getFloatVectorValues returned the bare dequantizing view OffHeapScalarQuantizedFloatVectorValues.

As a result, the scorer returned by those FloatVectorValues scores documents on the float values obtained by dequantizing the stored quantized vectors. This diverges from the normal path (full-precision vectors present), where the scorer computes scores directly on the quantized vectors.

The fix wraps the OffHeapScalarQuantizedFloatVectorValues with the quantized vector values, so that when full-precision vectors are absent, scoring uses the quantized vectors directly instead of first dequantizing them to floats and scoring on those. This mirrors the normal path: when full-precision vectors are present they are wrapped with the quantized values, and scorer() returns a quantized vector scorer.

TODO:

Add test case to simulate the behavior

[msokolov]

I think I understand the problem, but I confess the design of these classes is confusing, and I don't know how to evaluate whether this fix is correct. A test would help a lot!

[Pulkitg64]

Thanks @msokolov for taking a look. I agree this PR is incomplete without the test case. I will add a test case in next revision to simulate the failing behavior.