wip: start toying with wrapping requests withRedFmetrics

[ggilmore]

In addition to the metrics covered by RED, Google's SRE's "Monitoring Distributed Systems" recommends also capturing the the duration of failed operations as well. Tracking that separately seems important:

• the duration of failed operations can have wildly different timings to successful ones, so we shouldn't conflate the two in an average
• failed operations that are slow can still cause a bad UX

Zoekt seems to collect some info on the duration of failed requests, but it seems like it'd be nice to standardize that.

---

For now, this PR is toying with implementing the same pattern for bundling the RED metrics (+ the failed operation latency) that sourcegraph/sourcegraph's metrics package and influxDB uses.

It's not attempting to tie tracing/logging in the same module like sourcegraph/sourcegraph's metrics package - I thought it be better to keep it simpler for now.