Triggered task intercycle prereq bug

[MetRonnie]

Describe the bug
A triggered task that doesn't get submitted before the workflow shuts down, gets stuck on the previous cycle prerequisite on restart, causing stall.

Steps to reproduce the bug

[scheduling]
    cycling mode = integer
    runahead limit = P1
    final cycle point = 5
    [[graph]]
        P1 = foo[-P1] => foo
[runtime]
    [[foo]]
        script = """
            sleep 2
            if [[ "$CYLC_TASK_CYCLE_POINT" == "4" ]]; then
                cylc stop "$CYLC_WORKFLOW_ID"
                cylc trigger "${CYLC_WORKFLOW_ID}//2/foo"
            fi
        """

First run log:

...
INFO - Command succeeded: stop(mode=Clean)
INFO - Processing 1 queued command(s)
INFO - [2/foo waiting(runahead) job:01 flows:1] spawned
INFO - [2/foo waiting(runahead) job:01 flows:1] => waiting
INFO - [2/foo waiting job:01 flows:1] => waiting(queued)    <-------------------------- 2/foo not yet submitted
INFO - Command succeeded: force_trigger_tasks(['2/foo'], flow=['all'], flow_wait=False, flow_descr=None)
INFO - [4/foo running job:01 flows:1] (received)succeeded at 2022-07-27T16:05:30+01:00
INFO - [4/foo running job:01 flows:1] => succeeded
INFO - [5/foo waiting(runahead) job:00 flows:1] spawned
INFO - [5/foo waiting(runahead) job:00 flows:1] => waiting
INFO - [5/foo waiting job:00 flows:1] => waiting(queued)
INFO - Workflow shutting down - REQUEST(CLEAN)
INFO - DONE

Restart log:

...
INFO - Flows:
    flow: 1 (original flow from 1) 2022-07-27 16:05:02
WARNING - Partially satisfied prerequisites:
      * 2/foo is waiting on ['1/foo:succeeded']    <-------------------------------
CRITICAL - Workflow stalled

Expected behavior
2/foo doesn't wait for 1/foo:succeed, it should just run at restart because it was triggered.

Screenshots
(2/foo doesn't show up on restart)

Workaround

$ cylc set-outputs WORKFLOW_ID//1/foo

Pull requests welcome!
This is an Open Source project - please consider contributing a bug fix
yourself (please read CONTRIBUTING.md before starting any work though).

[hjoliver]

We'll need to store the force-triggered flag in the DB. (task.is_manual_submit)