Module quantization 

Quantization techniques for memory compression

This module provides tiered quantization strategies as specified in ADR-001:

Quantization	Compression	Use Case
Scalar (u8)	4x	Warm data (40-80% access)
Int4	8x	Cool data (10-40% access)
Product	8-16x	Cold data (1-10% access)
Binary	32x	Archive (<1% access)

Performance Optimizations v2

• SIMD-accelerated distance calculations for scalar (int8) quantization
• SIMD popcnt for binary hamming distance
• 4x loop unrolling for better instruction-level parallelism
• Separate accumulator strategy to reduce data dependencies

Structs

BinaryQuantized

Binary quantization (32x compression)

Int4Quantized

Int4 quantization (8x compression)

ProductQuantized

Product quantization (8-16x compression)

ScalarQuantized

Scalar quantization to int8 (4x compression)

Traits

QuantizedVector

Trait for quantized vector representations