Update box2d to v2.4.2

[oznogon]

Another go at #67 and #222, bumped to the final Box2D v2 release.

v2.4.1 introduced a breaking change to user data pointers, which this also attempts to incorporate.

[daid]

What are the benefits of updating? And the risks?

[oznogon]

Relevant to SP that I can spot:

• Several b2Polygon and b2PolygonShape performance and stability improvements, including fixes for a potential memory leak affecting non-circle Physics colliders (Fix memory leak in b2Polygon erincatto/box2d#486) and area calculation issues for polygons far from world origin (b2PolygonShape area calculation is zero when verts are large erincatto/box2d#422)
• Fixing static bodies that incorrectly and permanently awaken after a dynamic body is destroyed while in contact with the static body (Static body with contact, does not go to sleep after contacted dynamic body is destroyed. erincatto/box2d#450)
• Broad-phase/b2AABB performance improvements (Optimize broadphase AABB and pair management erincatto/box2d#589)

The only breaking change affecting SP that I could find was the change in user data using a struct member instead of a raw pointer. The other significant changes all appear to be related to net-new additions like linear shape casting, or fixes or changes in behavior for ropes, joints, and concave polygons, none of which I believe are used in EE/SP.

Or as someone smarter than me said about 2.4.1 in #67:

Daid: Not yet, might merge that now, not sure if it changes much. The changelog is petty small and most changes are in features not used by EE

I'm more concerned about potential issues with building on platforms that I don't build on, like MSVC.

[daid]

Well, mscv isn't an issue. And with all the ecs changes we can just pile this up as well I guess