Coming up with default histogram buckets for various metrics can be challenging, as we've seen in #274.

I'd like to propose some guidance for coming up with default histogram buckets which we can lean on in each time the issue of defining default histogram buckets comes up.

Proposal:

• If values align with service timings, use our default buckets, translated from millis to seconds, [ 0, 0.005, 0.01, 0.025, 0.05, 0.075, 0.1, 0.25, 0.5, 0.75, 1, 2.5, 5, 7.5, 10 ] (dropping the zero bucket?)
• Otherwise determine roughly the smallest and largest buckets that you care about, e.g. <0.01 seconds and >10 seconds, and use an exponential range in between (consider using base 2 if you need higher granularity and base 10 if you need lower granularity).

These would only be recommendations, and so if there is compelling reason to come up with a completely custom set of buckets for a particular metric that would always be ok.