smarter handling of filter duality

[gmbecker]

Right now it the collector checks whether it knows dplyr has been attached, but the problem here is that it is only checking in the current node, which often isn't where the library call appears.

Need to figure out a better solution for this.

[gmbecker]

There is a fix in place for this which tracks cumulative library loads across nodes when processing a whole script. We're still only checking for presence/absence of attaching dplyr, which isn't quite right (it could be masked later) but it's the best we're going to easily do so I'll wait for that to actually cause a problem before we try to tackle it further.