[BUG] Mojo crashes when invoking a parallel function calling a lambda without @parameter decorator

[p88h]

Bug description

Sample code below.

This tries to run a parallel worker with a (badly/non) parametrized lambda, which crashes mojo.
Should probably be detected as an issue at compile time?

Steps to reproduce

from algorithm import parallelize
from os.atomic import Atomic

fn main() raises:
    tiles = List[String]()
    tiles.extend(List[String]("abcde", "fghij", "klmno", "pqrst", "uvwxy"))
    dimx = tiles.size
    dimy = len(tiles[0])
    var mmax = Atomic[DType.int64](0)

    # this doesn't reference any globals so is ok
    fn idx(x: Int32, y: Int32) -> Int32:
        return 220 + ((y + 1) % 2) * 110 + x + 1

    # this references dimx, dimy but is not @parameter bound
    fn bar(i: Int) -> SIMD[DType.int32, 4]:
        return SIMD[DType.int32, 4](dimy, i - 2 * dimx - dimy, -1, 0)

    @parameter
    fn foo(start: SIMD[DType.int32, 4]) -> Int64:
        return int(idx(start[0], start[1]))

    @parameter
    fn step2(i: Int):
        mmax.max(foo(bar(i)))

    @parameter
    fn invoke() -> Int64:
        parallelize[step2](4)
        return mmax.value

    # this works, despite bar not having @parameter decorator
    print(foo(bar(1)))
    # and this causes the whole runtime to crash
    print(invoke())

    print(tiles.size, "tokens", dimx, dimy)

System information

- What OS did you do install Mojo on ?
Windows 11 / WSL 2
- Provide version information for Mojo by pasting the output of `mojo -v`
`mojo 24.5.0 (e8aacb95)`
- Provide Modular CLI version by pasting the output of `magic -V`
`magic 0.2.3`