While Bimpelrekkie's answer addresses how manufacturing process technology and transistor sizing influence voltage (and speed and power consumption), the business reasons for choosing a higher voltage are also worth considering.
In some cases, microcontrollers are designed to be drop-in (or near-drop-in, e.g., adding some connections with localized board changes) replacements. For the customer this can reduce design, validation, and even certification effort. The maximum operational voltage is then constrained by legacy requirements.
Supporting a higher voltage can also reduce the count and cost of system components outside the microcontroller. Motors, e.g., often use voltages higher than 3.3V; supporting more direct control of motors can reduce cost and increase reliability.
(Resilience under voltage variation can also be a factor in purchase decisions. Process technology, transistor sizing, and other factors influence voltage resilience.)
The choice of process technology is also guided by manufacturing cost. Using older process technology is less expensive per unit area; if pad count constrains area, using an older process is more cost effective (if energy efficiency and such are adequate). Also, for tiny integrated circuits, the wafer area lost from sawing may be significant, reducing the benefit of a smaller die. The availability of persistent memory (such as NOR flash) also influences process technology choice and biases toward the use of older processes.