Power efficiency and performance are main goals. Designing a logic family to operate with a well defined power supply range allows to optimize to meet those goals. In particular there is a distinct advantage in reducing (and thus limiting) power supply range, that is reduced power dissipation.
The now obsolete CD4000 CMOS family was indeed capable of handling a very large power supply range (the supply could range from ~5V to ~15V), but it was slow compared with more modern CMOS families.
Moreover, even in recent logic families there are special parts which allow operation with different power supply levels, if the need arise, such as those parts with open-drain outputs.
Keep in mind that logic chips are used essentially to perform "digital processing", and this doesn't require, in principle, high power levels (ideally one would want to use no power to process information). Whenever the need arise to drive devices that require a significant amount of power from logic level outputs you can always use standard interfacing techniques. For example, you can connect the logic output to a power MOSFET in order to switch on a relay. It would be wasteful (in many ways) to create an entire logic family capable of directly driving (relatively) high power loads when in reality this is only seldom necessary (in a complex digital board the number of "gates" that actually need to drive such loads directly is usually much smaller than the overall "gate count").