Flaky: NativeLinAlg.dot matrix-vector ComplexFloat returns wrong result on Linux

### Multik version

0.3.0

### Module

multik-openblas

### Platform

JVM

### Operating System

Linux

### Description

`NativeLinAlg.dot` for matrix-vector multiplication with `ComplexFloat` produces incorrect results intermittently on Linux (Ubuntu 22.04.5 LTS, gcc-12 12.3.0, JDK 21 Temurin). The test uses a 2x2 ComplexFloat matrix and a 2-element vector, asserting with epsilon `1e-4f`. Failed in 2 out of 4 CI runs.

### Reproducer

```kotlin
val matrix = mk.ndarray(
mk[mk[ComplexFloat(1, 2), ComplexFloat(3, 4)],
mk[ComplexFloat(5, 6), ComplexFloat(7, 8)]]
)
val vector = mk.ndarray(mk[ComplexFloat(1, 0), ComplexFloat(0, 1)])

val expected = mk.ndarray(mk[ComplexFloat(-3, 5), ComplexFloat(-3, 13)])

val actual = NativeLinAlg.dot(matrix, vector)
assertComplexFloatingNDArray(expected, actual, 1e-4f) // Fails intermittently
```

### Additional information

```text
Flaky — failed in 2/4 CI runs.
AssertionFailedError at NativeLinAlgTest.kt:270.
CI runs: 22511673326, 22511861936.
Environment: Ubuntu 22.04.5 LTS, gcc-12 (12.3.0), g++-12, JDK 21 Temurin, x86_64.
```

Start with NativeLinAlgTest.kt at the reported assertion around line 270 and the multik-openblas NativeLinAlg.dot entry point. Reproduce the 2x2 ComplexFloat matrix-vector case under the stated Ubuntu, gcc-12, and JDK 21 environment, then verify that the test passes consistently across repeated CI runs.

Written by the indexing model from the issue text.