There's a temporal aspect to your question, which isn't covered here. That is, which has better support now. That's quite distinct from which has better support in the future, or is likely to have.
The eGit project is, at the moment, still in development. It is possible to use for some things, but it's not self-hosting at the moment (in that it's still necessary to drop into the command line in order to make it happen). I have no such knowledge of the state of the Hg plugin.
However, Hg is licensed under the GPL and as such, won't be made part of an official Eclipse download ever. It doesn't matter how good (or bad) it is - the licensing prevents it from being used as part of an Eclipse.org download. And after all, one of the biggest problems that Eclipse has had with SVN for a while is the lack of a standard, out-of-the-box-it-just-works approach. That's why there's two competing processes, and even though one is an Eclipse.org project, it requires downloads from outside Eclipse in order to actually work.
At this point, the future of Eclipse and DVCS lies with eGit, whether it's good or not. It is quite likely that the Eclipse 3.6 series will have eGit support by default; and it's based on the same JGit library that NetBeans will use for the NetBeans implementation, so at least it's likely to be kept up to date.
There was a long debate about the merits of different DVCSs on Eclipse bug 257706 and the net result was for Git as the future DVCS for Eclipse, rather than other DVCSs. I wrote some other thoughts from EclipseCon at the time.
Disclaimer: I'm a contributor to eGit. I may well be biased, but I'm trying to report on the state of play of how likely it is for future releases of Eclipse to have DVCS support, rather than any statement of the current support for either.