optimise coalesce to fcoalesce

[jangorecki]

follow up of #3827

[MichaelChirico]

As I understand, we already saw the revdep impact of the provisional approach from #3800, so no need to bother with this until next release & we can start phasing it in then, right?

[jangorecki]

We could eventually drop this idea because of the differences in coalesce implementation described by Matt in #3754 (comment)
Unless BBmisc::coalesce would change its default behaviour, it seems to be more tricky. We will need to detect if BBmisc is on a search path. FYI @berndbischl

[jangorecki]

As commented, this doesn't seems to be good idea because of different implementations of coalesce. Best to just let people know they can use fcoalesce to have most standard and fastest implementation. Closing.