Fix job polling bug.

[hjoliver]

Manual cherry-pick 6ea4947 from #4437
Credit @wxtim

Close #4431

Bug reproduced on master with this workflow and time.sleep(10) between status file and job runner check in job_runner_mgr.py:

[scheduling]
    [[graph]]
        R1 = "poller & fast & waiter"
[runtime]
    [[poller]]
        script = cylc poll $CYLC_WORKFLOW_NAME fast.1
    [[fast]]
        init-script = sleep 5  # fake delayed start
    [[waiter]]
        script = "cylc__job__poll_grep_workflow_log '\(polled\)'"

Requirements check-list

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• Applied any dependency changes to both setup.py and conda-environment.yml (none)
• Does not need tests (note testable without bodging the code).
• No change log entry required (why? e.g. invisible to users).
• No documentation update required.

[dpmatthews]

Reproduced problem with Cylc 8 and confirmed this fixes it.

[wxtim]

@dpmatthews Any particular reason you didn't merge it?