contact: update modified time for better crash detection

[oliver-sanders]

Consider using a main loop plugin to update the modification time of the contact file at a scheduled interval for easier crash detection.

Problem

• When workflows crash, we are sometimes unable to remove the old contact file.
• The contact file acts as a marker for other Cylc commands (e.g. scan) that a workflow is running.
• In order to determine whether a workflow is running or crashed you need to ping the workflow.
• Pinging a workflow means a network request which is slow.

Proposed Solution

• Avoid the need to ping workflows by using the contact file modification time as an indicator of activity.
• os.utime can be used to set the modification time.
• No need for cylc scan to ping workflows UNTIL it detects crashes.
• This would only be local not remote.

See also #4858, #5405 (review)

[hjoliver]

Good idea, IMO.