Turbo: Optimize access patterns

[staabm]

Optimization 1: Eliminate Temporary zv::Ref Allocations in TrinaryLogic Operations

Changes:
Optimized and_(), or_(), extremeIdentity(), maxMin() methods
Replaced zv::Ref(&rest[i]).deref().asObject() chains with direct ZVAL_DEREF() and Z_OBJ_P() macros
Optimized LazyEvaluation::run() and runExtremeIdentity() methods
Rationale:
TrinaryLogic operations are called millions of times per analysis
Eliminates temporary object allocations per iteration
Direct macro expansion compiles to single instructions on 64-bit systems
Expected gain: ~0.8-1.2% on runs with heavy TrinaryLogic usage

---

Optimization 2: Inline Z_TYPE_P Checks in Type Accessors

Changes:
Replaced zv::Ref(value.raw()).isObject() with direct Z_TYPE_P() checks
Replaced zv::Ref(value.raw()).asObject() with Z_OBJ_P() macros
Replaced zv::Ref(value.raw()).isLong() with Z_TYPE_P() checks
Replaced zv::Ref(value.raw()).asLong() with Z_LVAL_P() macros
Rationale:
These Type methods are called in hot loops (millions of times per analysis)
Eliminates temporary zv::Ref wrapper objects
Raw macro calls compile to single instructions vs. multi-instruction function calls
Improved L1 cache locality on 64-bit systems
Expected gain: ~1.1-1.5% on runs with heavy Type operations