gnrc_rpl: takes unusually long time to start routing packets

[benpicco]

Description

Nodes may sleep for extended periods of time and schedule a wake-up time via RTC.
If they have their clocks synchronized they can all wake up at the same time, form a mesh and report their data / receive commands from the root node.

To increase battery life, the ON periods should be a short as possible.

gnrc_rpl takes an unknown and rather long time to settle. There is also no way to tell if the the mesh was successfully created (other than attempting to poll the root).

Steps to reproduce the issue

Run the test from #19134 but remove the arbitrary 5s sleep (time.sleep(5)).

Expected results

Routing packets to the RPL root should work immediately after the nodes have all received a valid global address / joined the DODAG.

Actual results

Routing does not work properly for ~3-4s after all nodes joined the DODAG and received a valid prefix.

Versions

RIOT master