* Set reachability table count to 1.
* Don't divide reachability write size by 4.
* Trim off all unused bytes after building reachability table.
* Default cvar 'navmesh_always_reachable' to 0.
AI behavior seems to have improved a bit, they attempt to pathfind better, and if they find a certain route they attempt another route where originally they would become stuck in such situation.
If you build the SDK after this commit, you need to rebuild all navmeshes, else the game would segfault in dtNavMesh::isPolyReachable().
dtNavMesh is now almost fully mapped out (most members should be correct, the only unsure ones are the unknowns and the mesh/tile flags).
dtNavMeshParams has one new field (I think the last field determines the precomputed malloc size, but I'm not very sure. It would make sense with the inline loop in 'Detour_LevelInit()' setting a certain field in each tile to -1).
dtMeshTile has one extra field (currently unknown) and the renamed unk0 field was is a pointer to the end of the 'polys' array.
The reachability table needs to be figured out still. The issue should be very small, but at the moment I do not have time for it.
The pointer to the table, and table pointers to data is correct, however, not a single poly is ever getting marked as 'reachable' (0xffffffff). This could be either within recast itself (see build_link_table() and set_reachable() functions), or the way the engine parses the data. The function that determines whether poly is reachable is located at '0x140F448E0'
