My choice is not only a shunt-wound motor, but a straight-shunt-wound motor. With the advent of SCR-based DC drive controllers, the need for specific motors has diminished. Especially in the case of a compound-wound, or a stabilized-straight-shunt-wound motor, the addition of a series field was an attempt to keep the motor under stable control during extended-speed operation, where the shunt field is weakened to attain speeds beyond base speed. They're also somewhat useful is containing armature reaction, though that's why motors are wound with inter-pole windings. But, they were designed to be used on old, constant-potential exciters (contactors stepping-through a bank of acceleration resistors, with the motor, and resistors connected to a regulated, fixed DC source).
Yes, motors featuring compound windings will afford a little more full-load torque, but the price that you pay is a very-unstable motor in the reverse direction. In cases where full-load torque is required in both directions, that series field will wreak havoc with stability, and commutation. Your only recourse at this point is to separately-excite the series field: both cumbersome, and expensive.
You don't need a series-wound motor to achieve stable, shared-load performance in a two-motor application. You can configure the master motor as a stiff speed regulator, and impose X-% of current droop on the slave motor. Current droop is nothing more than an over-summing of the velocity summing junction: I.E., 95-% velocity feedback, and 5-% current feedback. It will provide reasonably-balanced performance, and not require anything too exotic. The core disadvantage is that it also requires two DC drive controllers. Another, though more-difficult approach, is to drive both motors from one controller, and impose load-balancing resistors in the armature circuits. You may say that 2 series-wound motors accomplishes this effortlessly, but I'd never willingly-use a series-wound motor, because they require an applied load at all times. Even with a DC drive controller, performance will not be optimal, because you need a lot of armature current to develop a lot of torque. Some would argue that this is the case regardless, but in the case of a shunt-wound motor, you have full torque on-tap for the asking, plus decent no-load stability.
Finally, unless your budget can't afford it, never go with anything less than a 4-quadrant DC drive controller. You have absolute control of the motor, and don't have to worry about wearing-out directional power contactors.