MinBlockTimeTest hardening

[marta-lokhova]

This issue is tracking a couple of hardening tasks:

• Ensure fetching of the ledger age metric scales well when we run simulations with 100s of nodes. If k8s/supercluster adds additional long delays, this may skew the metric used to determine if SLAs pass or fail.
• Increase support beyond the default 5-minute window. Right now the metric tracks a 5-minute sliding window on the core side. Supercluster test happens to be 5 minutes as well, so we get an accurate result. We should harden this by scraping the metric every 5 minutes, which would allow us to run the test for longer (also, the current dependency between core and supercluster is a footgun anyways)

[marta-lokhova]

Tagging @drebelsky as well, since he's been looking into MinBlockTimeTest and might have feedback