v0.4.1

Multi-Camera Rendering

Added viewport fields (viewportX, viewportY, viewportW, viewportH) and clearFlags to Camera component
Per-camera render pipeline with viewport/scissor support for split-screen and minimap use cases
Camera priority-based render ordering
Editor scene view with per-camera color coding, camera icons, priority labels, and viewport preview overlay

Added BitmapText component for GPU-rendered bitmap font text with color, fontSize, alignment, spacing, and layer sorting
BMFont parser (.fnt text format) and LabelAtlas creation via the batch renderer
Editor support: font loading, inspector UI, thumbnail preview, click-to-navigate, and Tab navigation in BitmapFont glyph list

Spine runtime is now optional — none is the default when no Spine assets are used
Reduces build size for projects that don’t use Spine animation

Cached entityCount() with O(1) counter instead of O(n) scan
Replaced Camera/Canvas linear scans with C++ ECS view queries, eliminating ~4000 WASM bridge calls per frame
Reuse ext_vertex_storage_ buffers instead of clear+realloc each frame
Replaced O(n²) selection sort with std::sort in Registry::sort()
SpineRenderer: pre-allocate VAO/VBO/EBO, reuse with dynamic capacity
BatchRenderer: only bind active texture slots instead of all 8
Nine-slice: hoist texture slot lookup outside loop
TransformSystem: merge double iteration into single pass
SystemRunner: cache args array per system to avoid per-frame allocation