Use "X-Prometheus-Scrape-Timeout-Seconds" request header for timeouts

[Fra-nk]

Prometheus already tells us, which time it allots for the query by supplying the header X-Prometheus-Scrape-Timeout-Seconds (which defaults to the global scrape timeout and can be overridden on a per-job basis).
Prometheus' default configuration sets 10 seconds.

Aligning --timeout with the X-Prometheus-Scrape-Timeout-Seconds would simplify the general setup.

[jpeach]

Yep, we should pass the Prometheus scrape timeout along to the metrics snapshot timeout query parameter.