wow, that is a very dense mesh. If you are planning to use the skeletal animation and not just export image sequences that will be very heavy to run.
To answer your question. I would use far less vertices in the mesh, this will allow for the use of FFD to keep the volume of the legs when they bend. You can take a look at our second Kickstarter video to see the idea behind that https://www.youtube.com/watch?v=I5KbK-YxBKQ#t=200
Regarding the two different skinning modes Pharan is talking about they are commonly referred to as Linear Blend Skinning or Dual-Quarternion Blend Skinning.
Spine uses Linear Blend Skinning which does not keep the volume of a mesh when bending. The reason for not using Dual-Quarternion is that it is much heavier, for comparison Linear Blend Skinning runs as fast on CPU as Dual-Quarternion Blend Skinning does on GPU. I can't however say if it would be possible to have both methods implemented, probably not trivial.
In the future we do want to implement subdivision and it will likely be Catmull-Clark subd. This means a mesh does not get more manual control vertices but the new vertices have their position averaged out to give a smooth mesh, this allows for things like LOD where you can increase the subdivision at runtime if you need to have closeups of your character or lower it if you are far away or the framerate is low.