fix: Add factory methods to PythagoreanQuantizer

- Added for_llm(), for_embeddings(), for_vector_db(), hybrid() class methods
- Matches Rust core API for cross-repo compatibility
- All integration tests pass (15/17)

Verified:
- Hidden dimensions: k = ceil(log2(1/ε)) ✅
- Snap determinism ✅
- All quantization modes working ✅
- Lift/project round-trip ✅

Author: SuperInstance

constraint_theory/quantizer.py

@@ -244,6 +244,26 @@ def __init__(
         self._manifold = PythagoreanManifold(density=density)
         self._lattice = generate_pythagorean_lattice(max_hypotenuse=density)
 
+    @classmethod
+    def for_llm(cls) -> 'PythagoreanQuantizer':
+        """Create a quantizer optimized for LLM weights (ternary)."""
+        return cls(mode=QuantizationMode.TERNARY, bits=1)
+    
+    @classmethod
+    def for_embeddings(cls) -> 'PythagoreanQuantizer':
+        """Create a quantizer optimized for embeddings (polar)."""
+        return cls(mode=QuantizationMode.POLAR, bits=8, constraints=['unit_norm'])
+    
+    @classmethod
+    def for_vector_db(cls) -> 'PythagoreanQuantizer':
+        """Create a quantizer optimized for vector databases (turbo)."""
+        return cls(mode=QuantizationMode.TURBO, bits=4)
+    
+    @classmethod
+    def hybrid(cls) -> 'PythagoreanQuantizer':
+        """Create a hybrid quantizer that auto-selects mode."""
+        return cls(mode=QuantizationMode.HYBRID, bits=4)
+    
     def quantize(
         self,
         data: Any,