Exercise 1: A Three-Way Rotation (A -> B -> C -> A) — Possible Solution
====================================================================

THE TEST
------------------------------
    pa = kernel.create_process(num_pages=1)
    pb = kernel.create_process(num_pages=1)
    pc = kernel.create_process(num_pages=1)
    cpu = CPU()
    pa.transition(ProcessState.RUNNING)
    cpu.current_pid = pa.pid

    cpu.registers['PC'] = 100
    context_switch_fixed(cpu, pa, pb)   # A -> B
    cpu.registers['PC'] = 200
    context_switch_fixed(cpu, pb, pc)   # B -> C
    cpu.registers['PC'] = 300
    context_switch_fixed(cpu, pc, pa)   # C -> A

RESULT
------------------------------
    process A resumed after a full A->B->C->A rotation, CPU shows:
    {'PC': 100, 'ACC': 0, 'R1': 0, 'R2': 0}

A's own PC is correctly restored to 100 -- its exact value from before
it was ever switched away, even though it sat dormant through TWO
other processes' own turns (B running to PC=200, then C running to
PC=300) before finally being resumed.

WHY THIS GENERALIZES CLEANLY BEYOND TWO PROCESSES
------------------------------
context_switch_fixed()'s own logic never assumes there are only two
processes in the system -- it only ever knows about the ONE old
process and the ONE new process passed into a given call. Each call
is a fully self-contained save-then-load operation: A -> B saves A's
own state into pa.registers and loads B's saved state (all zeros, at
that point) into the CPU; B -> C saves B's own now-current state
(PC=200) into pb.registers and loads C's saved state; C -> A saves
C's own state (PC=300) into pc.registers and loads A's saved state --
which is still exactly what it was left as after the very first
switch, since nothing has touched pa.registers in between.

WHY THE OTHER TWO PROCESSES' OWN PROGRESS ALSO SURVIVES
------------------------------
Although this exercise only checks A's own final state, the same
mechanism guarantees pb.registers correctly holds PC=200 and
pc.registers correctly holds PC=300 at this point too -- each
process's own PCB is a genuinely independent storage location, never
shared or reused between processes. A three-way rotation doesn't
introduce anything qualitatively different from the two-way case; it
simply chains more individually-correct switches together.

WHY THIS WORKS AS AN ANSWER
------------------------------
Testing a rotation through a THIRD process, rather than just
alternating between the original two, rules out the possibility that
Finding 3's own correctness was specific to the two-process case --
confirming context_switch_fixed() is a genuinely general mechanism,
not a solution narrowly tailored to exactly two processes taking
turns.