In addition to Albert's note, it is worth noting that if we support LAG on the dataplane ports, you essentially eliminate redundancy. The whole purpose LAG is to increase bandwidth capacity, thus allowing to support 20Gbps using both 10Gbps ports as a LAG and not solely for the redundancy aspect.
Being that the SCG 200 is classified and mainly used for very large networks such as carriers and operators, it is essential that oversubscription is not allowed so that true redundancy can be accomplished when using both ports on a single dataplane interface. As such, this ensures that you always full 10Gbps capacity on each port even if one of the two ports fails (either via switch failure, physical transport failures or otherwise). The idea is that using a bridge allows us to ensure no more than 10Gbps is allowed through the port so that we have one other port able to support that.
The controller redundancy mechanism already performs load balancing between the two dataplanes (all four ports, used as bridge and as 2 interfaces). This means that if you have 10k APs as currently supported by the SCG-200 connected, and use both DPs, then you are essentially allocating 5k on each dataplane. Should one of the two dataplane fail, the failure would be detected and traffic or tunnels are moved to the other dataplane so that we can still support all 10k tunnels.
Hope this helps.