fix: update all Stage 2 references to Stage 3 CANONICAL type counts

Comprehensive update of type count references from Stage 2 (31 nouns + 40 verbs)
to Stage 3 CANONICAL (42 nouns + 127 verbs) across entire codebase.

Changes (23 files):
- Core architecture: Memory tracking comments, speedup calculations
- Tests: Type count assertions, enum index expectations, memory benchmarks
- CLI: User-visible type count output
- Augmentations: Type detection comments
- Documentation: Architecture docs, guides, performance docs
- Type embeddings: Regenerated for all 169 types (338KB)

Specific updates:
- 31 → 42 (noun count): 38 occurrences
- 40 → 127 (verb count): 24 occurrences
- 124 → 168 bytes (noun array size): 5 occurrences
- 160 → 508 bytes (verb array size): 5 occurrences
- 284 → 676 bytes (total type tracking): 12 occurrences
- Enum indices updated to match Stage 3 reordering

Type embeddings regenerated:
- 42 noun embeddings (64.5 KB)
- 127 verb embeddings (194.8 KB)
- Total: 338 KB (was 108.8 KB)

All constants, arrays, and tests now consistent with Stage 3 taxonomy.

Fixes #v5.5.1-type-count-migration

docs/PERFORMANCE.md

@@ -22,7 +22,7 @@ Where:
 - `k` = number of results returned
 - `m` = number of patterns to check
 - `f` = number of fields for entity type
-- `t` = number of types (30+ nouns, 40+ verbs)
+- `t` = number of types (42 nouns, 127 verbs)
 
 ## Architecture Deep Dive
 
@@ -140,7 +140,7 @@ The NLP processor uses **zero hardcoded fields** - everything is discovered dyna
 
 ```typescript
 class NaturalLanguageProcessor {
-  // Pre-embedded NounTypes (30+) and VerbTypes (40+) - ONLY hardcoded vocabularies
+  // Pre-embedded NounTypes (42) and VerbTypes (127) - ONLY hardcoded vocabularies
   private nounTypeEmbeddings = new Map<string, Vector>()
   private verbTypeEmbeddings = new Map<string, Vector>()
   
@@ -162,7 +162,7 @@ class NaturalLanguageProcessor {
 5. **Query Optimization**: Process low-cardinality type-specific fields first
 
 **Performance Characteristics:**
-- Type detection: O(t) where t = 70 total types (30 noun + 40 verb)
+- Type detection: O(t) where t = 169 total types (42 noun + 127 verb)
 - Field matching: O(f) where f = fields for detected type (typically 5-15)
 - Validation: O(1) lookup in type-field affinity map
 - No hardcoded assumptions - learns from actual data patterns