Performance optimizations and collision improvements

Performance:
- Remove expensive inverse() from all vertex shaders (terrain, WMO, M2, water, character)
- Add uniform location caching to avoid repeated glGetUniformLocation calls
- Add proper frustum culling for WMO groups using AABB intersection
- Add distance-based culling for WMO and M2 instances
- Add cleanup of unused M2/WMO models when tiles unload

Collision & Movement:
- Add M2 doodad collision detection (fences, boxes, etc.)
- Reduce character eye height (5.0 -> 1.8) and collision radius (2.5 -> 0.5)
- Enable WoW-style movement speed by default (14 units/sec run, 5 walk, 9 back)
- Fix emote grammar ("You waves." -> "You wave.")

Misc:
- Rename window title to "Wowee"

src/rendering/shader.cpp

@@ -92,7 +92,16 @@ void Shader::unuse() const {
 }
 
 GLint Shader::getUniformLocation(const std::string& name) const {
-    return glGetUniformLocation(program, name.c_str());
+    // Check cache first
+    auto it = uniformLocationCache.find(name);
+    if (it != uniformLocationCache.end()) {
+        return it->second;
+    }
+
+    // Look up and cache
+    GLint location = glGetUniformLocation(program, name.c_str());
+    uniformLocationCache[name] = location;
+    return location;
 }
 
 void Shader::setUniform(const std::string& name, int value) {
@@ -123,5 +132,9 @@ void Shader::setUniform(const std::string& name, const glm::mat4& value) {
     glUniformMatrix4fv(getUniformLocation(name), 1, GL_FALSE, &value[0][0]);
 }
 
+void Shader::setUniformMatrixArray(const std::string& name, const glm::mat4* matrices, int count) {
+    glUniformMatrix4fv(getUniformLocation(name), count, GL_FALSE, &matrices[0][0][0]);
+}
+
 } // namespace rendering
 } // namespace wowee