Better timeout handling for jobs

Right now we have timeout for kernel builds and LAVA jobs. But we do not check if process (k8s job, or LAVA job) is still alive. In some cases due large queue of jobs, we have situation when job is not started, but it is already marked as failed due timeout. We need to check if process is still alive, and if it is alive - extend timeout, if not - mark it as failed.

Start by locating the Python code that applies timeouts to kernel builds, Kubernetes jobs, and LAVA jobs, then trace how each job's running state is checked. The change is complete when a still-pending job has its timeout extended and a job whose process is no longer alive is marked failed.

Written by the indexing model from the issue text.