1.) For example PIC 16 and 18 families have their operands embedded in the instruction. Therefore each machine cycle must have a a instruction decoder before looking into the operand.
Most MCU systems decodes instruction by set of instruction decoders. They are simply a large collection of combinatorial circuits corresponding to each instruction.
It depends on the control unit and cpu architecture.for example in RISC architercture with pipeline each step work at the same time that others work that is each step calculate and do some work and send data to other steps. like a assembly section of a factory. each section do some work and send it to other section.but for example in serial architectures there is a counter that counts the steps and its output go to a decoder then in every step the decoder enables one section. if u want complete understanding about the cu and microcontrollers read computer orginigation written by Pattersone. or u can read computer architecture written by Moris mano. the latest is simpler then the first. at the end sorry for my bad writing.