Invert fill_tilemap usage of add_child to the tilemap parent entity to instead use add_children

[SirHall]

Please refer to and read through #375, decline this pull request if this is the wrong path.

[bzm3r]

This makes sense to me overall, but I you're right that we don't need to be adding the tiles as children of the tilemap in the first place. So maybe that will be the ultimate simplification?

[StarArawn]

This makes sense to me overall, but I you're right that we don't need to be adding the tiles as children of the tilemap in the first place. So maybe that will be the ultimate simplification?

I'm a little late to the party here, but we need to add tiles as children to the tilemap in order for despawn_recursive to work correctly or remind users that they need to despawn tiles. Otherwise this kind of thing can happen:
#384
Where you have dangling tiles(tiles no longer associated with a tilemap). This results in a panic in bevy_ecs_tilemap right now.

[SirHall]

Would you rather have all the fill methods add the tiles as children then?