Allow `overline()` to ensure that results are split by grouping variables and that the results can keep IDs

[Robinlovelace]

This issues is motivated by a real world need: nptscot/npt#17

Previous work:

• Mega thread on this that led to rnet_breakup_vertices(): rnet_breakup_vertices is way too slow for real network data #416
• https://docs.ropensci.org/stplanr/reference/rnet_join.html which can get to the same result, also see Test rnet_join() on large datasets #509

In this issue I aim to:

• Create an example that highlights the need
• Implement a solution in R (closing the issue)
• Describe the problem in a way such that the solution can be generalised