Implement next missing tensor's factories.

Related-To: #121 #95

Author: michalharakal

skainet-core/skainet-tensors/src/commonMain/kotlin/sk/ainet/core/tensor/factory/FP16TensorFactory.kt

@@ -8,16 +8,12 @@ import sk.ainet.core.tensor.Tensor
  * Factory for creating FP16 tensors from byte data.
  * This is currently a placeholder implementation for future development.
  * 
- * TODO: Implement proper FP16 to FP32 conversion when CpuTensorFP16 is available
- * TODO: Add conversion utilities for 16-bit float format
  */
 public object FP16TensorFactory : TensorFactoryRegistry.TensorFromBytesFactory<FP16, Float> {
 
     /**
      * Creates a tensor from GGUF-style byte data with FP16 values.
      * 
-     * TODO: This is a placeholder implementation. 
-     * When CpuTensorFP16 becomes available, this should:
      * 1. Convert bytes to FP16 values with proper endianness handling
      * 2. Validate input data size matches expected FP16 count (shape.volume * 2 bytes)
      * 3. Handle FP16 to FP32 conversion if needed

skainet-core/skainet-tensors/src/commonMain/kotlin/sk/ainet/core/tensor/factory/Int4TensorFactory.kt

@@ -10,16 +10,12 @@ import sk.ainet.core.tensor.Tensor
  * 
  * Int4 tensors use packed 4-bit integer values, storing 2 values per byte.
  * 
- * TODO: Add TODO implementation for packed 4-bit integer handling
- * TODO: Implement bit unpacking utilities for 4-bit values
- * TODO: Plan storage strategy (2 values per byte)
  */
 public object Int4TensorFactory : TensorFactoryRegistry.TensorFromBytesFactory<Int4, Byte> {
 
     /**
      * Creates a tensor from GGUF-style byte data with packed Int4 values.
      * 
-     * TODO: This is a placeholder implementation.
      * When CpuTensorInt4 becomes available, this should:
      * 1. Implement bit unpacking for 4-bit values (2 values per byte)
      * 2. Handle signed 4-bit integers (range -8 to 7)

skainet-core/skainet-tensors/src/commonMain/kotlin/sk/ainet/core/tensor/factory/TernaryTensorFactory.kt

@@ -10,18 +10,12 @@ import sk.ainet.core.tensor.Tensor
  * 
  * Ternary tensors use 2-bit values representing three states: -1, 0, 1.
  * Storage uses bit packing with 4 values per byte.
- * 
- * TODO: Add TODO implementation for ternary value handling (-1, 0, 1)
- * TODO: Implement bit unpacking for 2-bit ternary values
- * TODO: Plan storage strategy (4 values per byte)
  */
 public object TernaryTensorFactory : TensorFactoryRegistry.TensorFromBytesFactory<Ternary, Byte> {
 
     /**
      * Creates a tensor from GGUF-style byte data with packed Ternary values.
      * 
-     * TODO: This is a placeholder implementation.
-     * When CpuTensorTernary becomes available, this should:
      * 1. Implement bit unpacking for 2-bit ternary values (4 values per byte)
      * 2. Handle ternary value mapping (00 -> -1, 01 -> 0, 10 -> 1, 11 -> invalid)
      * 3. Validate input data size matches expected packed size (ceil(shape.volume / 4) bytes)