MGS could record a metric for the number of crash dumps on each SP

[hawkw]

@mkeeter points out to me that it could be quite useful for the MGS metrics subsystem to report a counter metric tracking the number of task crash dumps present on a SP. This would allow us to easily see which SPs on the rack have crash dumps, as well as providing an indication of the time¹ at which a new crash dump appeared, which provides an approximation of when the task crash occurred.

Footnotes

1. Wall clock time as understood by MGS. ↩

[hawkw]

This would probably be recorded under a new target, rather than the hardware_component target used by sensor metrics, as it's scoped to the whole SP rather than a particular hardware device known to the SP.