fix reversely PC wrap problems

[linkaishengmri]

Problem

When the PC is paused (e.g., due to a wait instruction), and the old_pc_hw value is 0, the next pc_hw value may wrap backward by 4 bytes (e.g., from 0x0 to 0x3fffc), due to hardware behavior. This inverse wrap was not correctly handled in the code, and pc_offset remained unchanged.

As a result, the program counter (pc) appeared to advance incorrectly by 0x40000, causing pc > mem_offset, and ultimately triggering a memory buffer underrun.

Fix

Added the following check in the PC tracking logic to detect and correct inverse wrap-around:

if (pc_hw > old_pc_hw + MARGA_MEM_SIZE - 16) {
    // PC has jumped backward due to a wrap from waiting/pausing 
    debug_printf("PC Inversely wrapped: Loop %u | Mem: 0x%zx | RxReads: %u | Exec: %u | PC_HW: 0x%zx | OLD_PC_HW: 0x%zx\n",
        execution_loops, mem_offset, rx_reads_per_loop, exec, pc_hw, old_pc_hw);
    pc_offset -= MARGA_MEM_SIZE;
}