Investigate problems with uptime analysis for ITN.

[Sventimir]

The uptime service backend is responsible for recording and analysing the uptime submissions from nodes. The analysis script tends to produce suspicious (or outright wrong) results lately. Besides it is also very tightly coupled to AWS and Google infrastructure, which makes iot hard to test. We need to improve the testability and resolve the bad results issue. Additionally at the moment the service distinguishes submissions from different IP addresses, even if they're signed by the same key. From now on we prefer to ignore IP addresses and count all submissions from a public key for the same identity.

[Sventimir]

For the moment I managed to improve testability by decopupling from GCP spreadsheet and outputting results to stdout instead. Ongoing work is available here.