Regarding point 1:
To be honest I am not 100% sure. However, here are my thoughts:
As with all profiles the turning of the flow is easier when it is accompanied with a pressure gradient (from high to low pressure). Easier is quite an unspecific word lumping together a lot of undesirable aerodynamic features like boundary layer growth and or boundary layer separation. The gradient between high pressure and low pressure can be reduced by increasing blade length (reducing the pressure-gradient acting on the boundary layer).
Regarding your mention of subsonic flow and cross section: In the case of a transonic compressor the inflow is supersonic and the outflow is subsonic. Plus, the compressor flow is always from a lower to a higher radius (the turbine is the other way around). So both sides will feature a bigger area at the higher radius and a smaller area ant the smaller smaller radius.
So I guess we look at the compressor with the turbine in the background.
Because it features longer blades with reduced flow turning at the outer diameter and splitter blades (end-bend).
Regarding point 2:
The design of a radial compressor is influenced by a lot of parameters one is manufacturability. I guess that the requested pressure rise of the compressor and the associated turning of the flow was not possible with just seven blades, however the machining and minimum blade thickness did not allow for 14 full blades. If you are able to provide a meridional view of pressure rise one would probably see that the first section of the compressor has a lower loading and the rear sections of the passage feature a higher loading because it can be distributed across a larger number of blades.
So in short the answer is: The design depends on the requirements in terms of manufacturability and thermodynamic behaviour (e.g. off-design characteristics). And what we see seems to be the optimal trade-off.
Regarding point 3:
You need to keep in mind that NACA profiles are based on constant free stream flow velocity. In an axial machine this assumption is somewhat valid because the difference in radii of inflow and outflow stream surface are
small. In an radial machine the free stream velocity changes due to the higher circumferential velocity (i.e. higher radius).
NACA profiles do not account for centrifugal forces which play a huge role in radial machines.