What you're talking about is the rise of specialization, which is to say, at what point do various disciplines develop to a point where it becomes unlikely that any single person can be a true master of multiple, different ones that require different skill-sets. And the answer for this is that there isn't some sort of "cut-off" date or anything, but by the late 19th century science had "professionalized" to a degree that rapid specialization was occurring, and in the 20th century this was pushed to sort of exponential heights as science itself became a much larger and better-supported enterprise.

So in the 17th century, a very clever person like Hooke or Newton could, with more or less a single "tool set," tackle a huge variety of scientific problems. For Hooke the "tool set" was very keen observational skills; for Newton it was advanced mathematics. Through both of these applications they conquered quite a lot of "low-hanging fruit."

By the late 19th century, though, you'd need far more than observational acuity to make significant, novel contributions to physics, biology, anatomy, and mathematics. And advanced mathematics was quickly becoming a prerequisite for participation, and less of a unique advantage.

In terms of "averages," though, and not outliers, I suspect it would come down to how one defined "averaged" and who one decided it was. Being specialized and capable of contributing novel research at even one thing is very difficult, as anyone who has gotten a PhD can tell you. Being able to contribute to more than one thing in major, significant ways, is rare at any time in history. I think one can say that the average scope of a scientific education has changed over time — specialization has increased — so by some measures the "averages" are more specialized in the past, but I don't think one could really claim that the "average" was more polymathic in the past beyond the fact that the disciplines had inherently wider borders.

In terms of evidence: there are actually interesting and newly-pursued quantitative/computational methods to characterize how a given scientific discipline "works" over time, and how much range a given scientist or collection of scientists had. So one can, through citation analysis and topic analysis, feed journal articles through programs that will generate network diagrams of their authors and subjects, and allow you to view how those diagrams evolve over time. You can read about some of these methods here. This work is on-going. I am not sure if they have looked at this specific question yet (their focus so far are individual disciplines and how they change and split, and widening that to include multiple disciplines would be tricky), but one could use similar methods. More qualitatively, one can indeed just read the proceedings of scientists beyond the "outlier" ones — it is a common exercise to ask students, for example, to pick a random issue of an older scientific journal (say, The Physical Review, 1901) and use the articles therein to characterize the way the field is working at that time.

For what it is worth, I would not consider Feynman in this category. His major contributions were in theoretical physics, fairly narrowly. This in no way takes away from his cleverness to point this out. (Though for what it is worth, he himself thought his contributions were relatively minor compared to the works of others. His friend/frenemy Murray Gell-Mann's contributions were much deeper, but Gell-Mann wasn't as good as self-publicist, so most people don't know about him at all.) Von Neumann did have a wide range, but specifically because the mathematics that he practices applied to many different contexts. Among the physicists of the "Golden Age" of the mid-20th century, Fermi was often credited as the most wide-ranging in his capabilities by his peers, being a capable experimentalist and a capable theorist. The fact that such was considered a "wide-range" by the 20th century — two different modes of the same discipline — gives some further evidence to how much these categories had contracted!

As for whether people have a "great man" view of scientists — yes, of course. Historians of science today put very little credit on individual "great men" pushing the field forward; they tend to take a more sociological, community-based approach. Science is not one individual, it is an interconnected community of people both inside and outside of the scientific community. Even when it appears to be one "great" person (a Galileo, a Newton, a Darwin, an Einstein), looking a little deeper shows these people reliant on, and engaging with, much larger networks of interlocutors, and the rhetorical power of their work had to be amplified by others as well. So ultimately it reduces down very similarly to the "great men in history" issue: you cannot understand the individual without the context, and the work of the historian is frequently to understand both simultaneously. Many biographical works of the history of science are now built around this idea; see e.g. Browne's biographies of Darwin, Galison's biography of Einstein; Westfall's biography of Newton; Biagioli's biographies of Galileo; etc.