Improve shadow performance: halve resolution, 9x fewer PCF taps, throttle depth pass

- SHADOW_MAP_SIZE 2048→1024: 4x fewer pixels rasterized in depth pass
- Replace 9-tap manual PCF loop with single hardware PCF tap in all 4 receiver
  shaders (terrain.frag, wmo_renderer, m2_renderer, character_renderer).
  GL_LINEAR + GL_COMPARE_REF_TO_TEXTURE already gives 2×2 bilinear PCF per
  tap for free, so quality is maintained while doing 9x fewer texture fetches.
- Throttle shadow depth pass to every 2 frames; OpenGL depth texture persists
  between frames so receivers always have a valid shadow map. 1-frame lag at
  60 fps is invisible.

src/rendering/m2_renderer.cpp

@@ -391,14 +391,8 @@ bool M2Renderer::initialize(pipeline::AssetManager* assets) {
                         float edgeDist = max(abs(proj.x - 0.5), abs(proj.y - 0.5));
                         float coverageFade = 1.0 - smoothstep(0.40, 0.49, edgeDist);
                         float bias = max(0.005 * (1.0 - abs(dot(normal, lightDir))), 0.001);
-                        shadow = 0.0;
-                        vec2 texelSize = vec2(1.0 / 2048.0);
-                        for (int sx = -1; sx <= 1; sx++) {
-                            for (int sy = -1; sy <= 1; sy++) {
-                                shadow += texture(uShadowMap, vec3(proj.xy + vec2(sx, sy) * texelSize, proj.z - bias));
-                            }
-                        }
-                        shadow /= 9.0;
+                        // Single hardware PCF tap — GL_LINEAR + compare mode gives 2×2 bilinear PCF for free
+                        shadow = texture(uShadowMap, vec3(proj.xy, proj.z - bias));
                         shadow = mix(1.0, shadow, coverageFade);
                     }
                 }