Incorrect prologue for a while loop over a list of tensors

🐛 Bug

I was preparing an example reproducer for one problem and hit the following error:

File ~/dev/lightning-thunder/thunder/__init__.py:623, in jit.<locals>.get_computation_and_inputs(*args, **kwargs)
    620 cs.last_interpreter_log = last_interpreter_log
    621 cs.last_interpreted_instructions = (i for i in last_interpreter_log if isinstance(i, dis.Instruction))
--> 623 inps, pro_to_epi = pro(*args, **kwargs)
    625 computation_trc = dce(computation_trc)
    626 computation_traces.append(computation_trc)

File ~/dev/lightning-thunder/thunder/__init__.py:727, in jit.<locals>.prologue_execution_timer.<locals>.wrapped(*args, **kwargs)
    725 cs.last_prologue_execution_start = time.perf_counter_ns()
    726 try:
--> 727     return fn(*args, **kwargs)
    728 finally:
    729     cs.last_prologue_execution_stop = time.perf_counter_ns()

File thunder.prologue_2:10, in prologue(*args, **kwargs)
      8 check_len(kwargs, 0)
      9 subscr: "Any" = args[0]
---> 10 t0: "cpu f32[2, 2]" = subscr[1]
     11 a: "cpu f32[2, 2]" = args[1]
     12 t_0_0: "cpu f32[2, 2]" = subscr[0]

IndexError: list index out of range

Sample script to trigger the error:

import torch
import thunder

def func(t0s, a):
    t1s = []
    while t0s:
        t0 = t0s.pop()
        t1 = t0 + a
        t1s.append(t1)
    return sum(t1s)

jfunc = thunder.jit(func)
t0s = [torch.randn(2, 2) for _ in range(2)]
a = torch.randn(2, 2)
jfunc(t0s, a)

Start with the sample script and inspect thunder/init.py around the prologue invocation at lines 623 and 727, then examine the generated thunder.prologue_2 shown in the traceback. Reproduce the list-consuming while loop and compare the generated indexing with the loop's execution; done means the script runs without IndexError and returns the expected tensor result.