Simplify README optimization section - remove PGO

Remove Profile-Guided Optimization from README.md as it's not practical
for a library with unpredictable workload patterns:

- PGO training data is inherently biased to specific use cases
- SQL parsing workload varies wildly (dialects, query complexity, patterns)
- Adds significant build complexity (3-phase build) for marginal gains
- Modern compilers with LTO already optimize effectively

Replaced with practical recommendations:
- Architecture-specific builds (-march=native) for 5-15% gains
- Simplified compiler optimization explanation
- Moved PGO to OPTIMIZATION_STRATEGY.md as "advanced, specialized only"

Focus README on optimizations that benefit all users, not edge cases.

🤖 Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)

Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>

Author: 3 people

README.md

@@ -221,22 +221,20 @@ ctest --test-dir build
 
 ### Advanced optimisations
 
-**Profile-Guided Optimisation (PGO)**: For production deployments requiring maximum performance, enable PGO in 3 steps:
+**Compiler optimisations** (Release builds): Link-Time Optimisation (LTO), aggressive inlining, constant folding, constant merging, symbol visibility optimization. All enabled by default with `-DCMAKE_BUILD_TYPE=Release`.
+
+**Architecture-specific builds**: For maximum performance on your specific CPU, build with native architecture optimizations:
 
 ```bash
-# Step 1: Build with profiling instrumentation
-cmake -B build -DCMAKE_BUILD_TYPE=Release -DLIBSQLGLOT_PGO_GENERATE=ON
+cmake -B build -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS="-march=native"
 cmake --build build
+```
 
-# Step 2: Run with representative workload to collect profile data
-./build/benchmarks/bench_transpiler  # or your own queries
+This enables CPU-specific instructions (AVX2, AVX-512, etc.) for your exact processor, typically 5-15% faster than generic builds.
 
-# Step 3: Rebuild using profile data for optimisation (10-30% faster)
-cmake -B build -DCMAKE_BUILD_TYPE=Release -DLIBSQLGLOT_PGO_USE=ON
-cmake --build build
-```
+**Benchmarking**: Comprehensive benchmark suite available. Build with `-DLIBSQLGLOT_BUILD_BENCHMARKS=ON` to measure performance on your workload.
 
-**Compiler optimisations enabled** (Release builds): Link-Time Optimisation (LTO), constant merging, symbol visibility optimisation.
+See `docs/OPTIMIZATION_STRATEGY.md` for profiling tools, per-file optimization levels, and advanced techniques like Profile-Guided Optimization (useful only for specialized, consistent workloads).
 
 ## Architecture
 

TESTING.md tests/TESTING.mdTESTING.md renamed to tests/TESTING.md

@@ -201,29 +201,29 @@ docker compose -f docker/docker-compose.yml run --rm test
 | **Feature Combinations** | 3,150+ |
 | **Pass Rate** | 100% |
 
-## Code Quality Standards
+## Code quality
 
-### Memory Safety
-✅ Arena allocator with:
+### Memory safety
+Arena allocator with:
 - Integer overflow protection
 - Maximum allocation limits (1GB)
 - Alignment safety checks
 - RAII with `std::unique_ptr`
 
-✅ No manual memory management:
+No manual memory management:
 - No `malloc/free`
 - No `new/delete` (except placement new in arena)
 - No unsafe C functions (`strcpy`, `strcat`, `sprintf`)
 
-### Parser Safety
-✅ Bounds checking on all token access:
+### Parser safety
+Bounds checking on all token access:
 ```cpp
 void advance() {
     if (!is_at_end()) pos_++;  // Safe advancement
 }
 ```
 
-✅ Null pointer checks before string construction:
+Null pointer checks before string construction:
 ```cpp
 if (is_at_end() || current().text == nullptr) {
     break;
@@ -292,17 +292,6 @@ When adding dialect-specific features:
 - [ ] Continuous fuzzing integration
 - [ ] Dialect compliance testing against vendor documentation
 
-## Contributing
-
-When submitting code:
-
-1. ✅ **All 492 tests must pass**
-2. ✅ Add tests for new features
-3. ✅ Add fuzzing tests for edge cases
-4. ✅ Update this document
-5. ✅ Follow security best practices
-6. ✅ Maintain code quality standards
-
 ## Continuous Integration
 
 Tests run automatically on: