Fix fetch certs interval

[sea-snake]

As observed by @peterpeterparker, earlier retry delays only applied to the first outcall error instead of any failed future outcalls.

Thanks @peterpeterparker for catching this one.

Changes

• Switch around meaning of delay argument in schedule_fetch_certs, None now default to FETCH_CERTS_INTERVAL instead of 0.
• When outcall succeeds, schedule_fetch_certs is called with None:
  • Will default to FETCH_CERTS_INTERVAL with timer.
  • Will use 60 as retry delay.
• When outcall fails, schedule_fetch_certs is called with Some:
  • Will use delay with timer.
  • Will use min(FETCH_CERTS_INTERVAL, max(60, delay * 2)) as retry delay.

Tests

Added should_compute_next_certs_fetch_delay test to make sure delay is computed correctly.

[lmuntaner]

Nitpick but already approving.

[lmuntaner]

Thanks!

[sea-snake]

@lmuntaner Also renamed to MAX_VALIDITY_WINDOW to MAX_VALIDITY_WINDOW_SECONDS. Keep in mind that it was previously nanoseconds, had to update it's usage in two places with secs_to_nanos().