Update metrics collector to track readiness

In some cases it may take a lot of time between when the instance provisioned by fleeting enters the running state and when it's really ready to be used by taskscaler. This may change further depending on what taskscaler's caller wants to do.

This MR adds a metric that tracks timings of the instance becoming ready (counted from when it's discovered to be in the running state). Including the optional caller's upFunc() execution.