I think I'd like to add some more clarification to the nature of plugins. In my mind, anything beyond "what makes it easy to edit text of any kind" should be put into the plugin category. Syntax files help "make it easy to edit text of any kind", so they can stay, as can features like spaces-to-tabs, search/replace, etc... Everything else is a plugin. FTP, CVS and SVN support should be plugins, for example, as they're more specific tools.
So that's the first and more important criteria. The second criteria would obviously be popularity. Just because it helps it make it easy to edit text of any kind, does not mean it should automatically be built-in. i.e. If its not very popular then it should be made a plugin.
Given those two prioritized criteria, I should point out that just because something is relegated to a plugin, does not mean the dev team shouldn't build it. If the dev team thinks a feature is popular enough, they should take it upon themselves to build an official plugin to do the job.
As for installing plugins, it should be a simple enough task for end-users to find/download/install plugins. I'm envisioning a system similar to Firefox Extensions or even the SMF Forum Mods.
The last thing is, while I think popularity should be considered secondary, if there are performance losses associated with making a feature a plugin, then perhaps popularity needs to actually be the most important criterion. That is, if a feature will not be as fast or efficient as a plugin compared to being built-in, and that feature is also very popular, then even if its not directly related to "making any kind of text easier to edit", it makes sense to build-it-in.