Improve fw update (more generally cron) resilience

[xmo-odoo]

The forwardport.queue model builds in a latched retry behaviour to avoid spinning endlessly, however when github is in a bad way that has a few issues:

• it spins (triggers) as fast as possible, so if there's little to no queue the error condition is unlikely to have resolved
• it never signals that it has latched, so discovering that that happened can be difficult
• resetting the latch is a pain in the ass, the latch info is not displayed anywhere and there's no easy action to reset the latch

Also because I forgot about this mechanism (?) the main forward port queue has a separate but overlapping latching system, with mail.thread integration.