share/discovery: overly aggressive retrying

I was looking into the mocha network today and found out a few things we need to change in discovery:
* We retry too often.
* If we cannot get the needed value of peers, we retry every second. Instant retries are usually useless; no new peers would join the network or advertise themselves in the second timeframe. But they are not always useless, so
* Instead, we should have a capped linear backoff.
* Factor 5 starting from 5 seconds and capped by 10 mins
* #2573 introduced a minor bug: `discoveryLoop` does not listen to `triggerDisc` channel anymore
* We should aim to trigger `discoveryLoop` asynchronously. RIght now, we endlessly trigger the loop every second, even if we have enough peers. Instead, we should rely more on `triggerDisc` channel to reduce wasted cycles. (nice-to-have)