Fix Issue #82 - Pushed boulder can collide with pusher

[crkellen]

This is a fix for #82.
The issue is twofold:

1. Boulders pushing movement is calculated expected a perfect resolution. It uses truncation (floor()) in it's calculations, which cause it to be fixed to grid positions (only moves in increments of 16). This is only a problem when the boulder collides with something like another boulder (or other Entity), because the boulder will be somewhere in between the increment of 16. When a push happens after that, the floor() calculation causes the boulder to jump back to it's correct position (starting position). The player may be standing within that starting position, which causes the boulder to collide and hurt the one pushing.

2. Calculations for collisions with boulders is done using 'entityInsideEntity()'. This is a simple AABB bounding box collision check, which allows for a collision to occur from the opposite direction than the boulder is rolling. The solution is to implement 'entityInFrontOfEntity()' which, for each direction, avoids the possibility of colliding in the direction opposite of movement.

So, the collision is completely fixed, however there are still a few things left to sort out with boulders:

• Clients do not have accurate positions of boulders (constantly vibrating if not against a wall/entity)
• Boulders have a size that is 1 unit less than their actual radius. This allows for them to be partially inside of another Entity, such as another boulder, which causes a visual issue of the boulder jumping back when the floor() calculations are done.

These will be solved in future pull requests.

[lheckemann]

A lot of new duplication against boulderCheckAgainstEntity. Indeed, the functions are identical besides the distinction between entityInFrontOfEntity/entityInsideEntity. Common functionality should be factored out.

But I am really looking forward to not dying in Sokoban xD

[crkellen]

I agree, I guess I didn't consider that after finally figuring out the issue. All fixed now!

[lheckemann]

Much nicer! I haven't tested it though so this is only approval of the code itself.