Bug in run_cmd_with_timeout means you always get an error about `ls -d` timing out

Nobody has claimed this yet.

If the `SCRATCH_ROOT` env var is set you hit this code: https://github.com/lowRISC/dvsim/blob/0f1346640562da32bc399530b309487e9e1df832/src/dvsim/cli/run.py#L62-L72

That calls this function: https://github.com/lowRISC/dvsim/blob/0f1346640562da32bc399530b309487e9e1df832/src/dvsim/utils/subprocess.py#L28-L39

(With the wrong type for `exit_on_failure` btw, but who needs correct types?)

The problem is that function will basically *always* say that the command has timed out. The bug is due to the classic truthiness footgun that Python shares with Javascript. Unlike in Typescript where you can (with considerable effort) set up ESLint to ban truthiness bugs, I'm not aware of a way to do that in Python so you just have to be vigilant.

Here is the bug (and the other `poll` further down): https://github.com/lowRISC/dvsim/blob/0f1346640562da32bc399530b309487e9e1df832/src/dvsim/utils/subprocess.py#L53-L54

It needs to be `if p.poll() is not None:`.

`poll()` returns the exit code or `None` if the process has not exited. If the command succeeds then `p.poll()` returns 0, and `if 0:` will never succeed so it thinks the command times out.

Start with src/dvsim/cli/run.py and the run_cmd_with_timeout implementation in src/dvsim/utils/subprocess.py, following the SCRATCH_ROOT path and the referenced poll checks. Verify the behavior with the `ls -d` command and confirm that a successful exit is no longer reported as a timeout.

Written by the indexing model from the issue text.