infer: an infinite loop over a spy grows the trace until the memory is gone #873

Split off from #774, which covers the sys.settrace mechanism. This one is an infinite loop over a spy:

import multiprocessing.heap
infer(multiprocessing.heap.Heap._free_pending_blocks)

while True:
    try:
        block = self._pending_free_blocks.pop()
    except IndexError:
        break
    ...

self._pending_free_blocks.pop() on a spy never raises IndexError, so the loop runs until the 60 s isolate timeout. Every iteration appends to the trace, so memory grows at roughly 4 GB per 12 s: under a 4 GB RLIMIT_AS it ends in InferError: the function never ran to completion (MemoryError), without a cap it exhausts the machine first.

Nothing bounds the number of operations a single run records; the run, fork, and yield budgets do not apply to a loop that never forks or yields. A per-run operation cap (raised as a BaseException so the target cannot swallow it, and reported as an exploration gap) would end this class.