NDArray M6: Int/Boolean/Float element types with ops, reductions, boolean indexing, and where

[Copilot]

• Part 0: Fix NDArray[Double] axis argmax/argmin return type to NDArray[Int]
• Update existing reduction tests for the type change
• Part 1a: ndarrayIntOps.scala — NDArray[Int] element-wise ops
• Part 1b: ndarrayIntReductions.scala — NDArray[Int] reductions
• Part 2: ndarrayBooleanOps.scala — NDArray[Boolean] logical ops + reductions
• Part 3a: ndarrayFloatOps.scala — NDArray[Float] element-wise ops
• Part 3b: ndarrayFloatReductions.scala — NDArray[Float] reductions
• Part 4: ndarrayBooleanIndexing.scala — boolean mask indexing
• Part 5: ndarrayWhere.scala — conditional where
• Part 6: all.scala exports updated
• Test files: Int ops, Int reductions, Boolean ops, Float ops, Float reductions
• Fix native compile error: Removed import vecxt.BooleanArrays.* from ndarrayBooleanOps.scala; replaced a.data.not, a.data.any, a.data.allTrue, a.data.trues, a.data.not! delegations with inline while loops to avoid local extension name shadowing
• Fix native compile error: Added @targetName annotations to all NDArray[Int] and NDArray[Float] extension methods that conflict after type erasure when exported to all.scala (same method names as NDArray[Double] extensions — product, sum, mean, min, max, argmax, argmin, neg, abs, +, -, *, /, >, <, >=, <=, =:=, !:=, +=, -=, *= etc.)
• Test files for boolean indexing and where (not created)

[Quafadas]

@copilot can you check the failiing tests on native ?

[Copilot]

@copilot can you check the failiing tests on native ?

Fixed in e20d97b. Two issues on native:

1. BooleanArrays scope conflict — import vecxt.BooleanArrays.* inside NDArrayBooleanOps caused the local any/not/not! extension methods on NDArray[Boolean] to shadow the imported extensions on Array[Boolean]. Fixed by removing the import and replacing all delegations (a.data.any, a.data.allTrue, a.data.trues, a.data.not, a.data.not!) with direct inline while loops.

2. Type-erasure "Double definition" conflicts — Extension methods on NDArray[Int] and NDArray[Float] share the same erased type signatures as the existing NDArray[Double] methods (e.g. product, sum, >, neg, etc.) when all are exported to all.scala. Scala Native is strict about this. Fixed by adding @targetName annotations to all conflicting extension methods in ndarrayIntOps.scala, ndarrayIntReductions.scala, ndarrayFloatOps.scala, and ndarrayFloatReductions.scala.