bug(trainer): ContainerBackend.wait_for_job_status missing polling_interval validation

## What happened?

`ContainerBackend.wait_for_job_status` does not validate that `polling_interval`
is less than `timeout`. Both the kubernetes and localprocess backends have this
guard, but the container backend is missing it entirely.

When `polling_interval >= timeout`, the time-based while loop exits before
the first sleep completes — job status is checked zero times — and a
`TimeoutError` is raised immediately with a misleading message.
No error is raised for clearly invalid input.

## What did you expect?

A clear `ValueError` should be raised before the loop starts, matching
the behavior of the other two backends.

## Affected file

`kubeflow/trainer/backends/container/backend.py` — `wait_for_job_status`

## Related

Same class of bug as #400 (kubernetes backend), fixed in #402.

Start in kubeflow/trainer/backends/container/backend.py at ContainerBackend.wait_for_job_status, then compare the polling_interval validation in the kubernetes and localprocess backends and the related fix in #402. Done means invalid intervals raise ValueError before the polling loop, matching the other backends.

Written by the indexing model from the issue text.