Is there a technical reason you're asking as a dev?
That data is where the gold is buried, so... I do recall seeing certain #s in the Annual Report tho.
As far as sharing w/devs, Apple tends to stick with iOS versions, and platforms, not so much specific hardware, see: https://developer.apple.com/support/app-store/
Otherwise, google is your friend - that, and any willingness you may have towards funding a white paper.
I did try googling it, but only got third-party stats, especially from one guy who published the data from his audiobook app installs.
But I guess you're right, aside from the annual reports, I don't see any other hard source.
In response to your edit, yes, I think there is a technical reason.
Call me old-fashioned, but I've yet to trust emulated devices to fully behave like physical ones. Coming from an Android background, especially regarding Google/Apple services. but I also had trouble using emulated Watches with emulated phones (a no-go in Android, actually), and physical sensors, especially barometrics.
I believe it's better to test on a physical device using the debugger, and if I'm to buy an iPhone for development, I'd want the one with the biggest market share (the most "average", if you will) to give a good picture about my target audience.
I agree with device testing. Sim testing is ok for exercising an app's UI, but little trust should be given beyond that, I think - nothing old fashioned in that logic. Anyone that says sim testing alone makes an app good-to-go is naively asking their users to test for them...ouch.
As for device testing v sample pool, I'd go for width, over singular depth. Rather than hope to target the single deepest demographic on one particular device* with a single biggest 'market share', I prefer device diversity. If you're talking about trusting your device testing, covering your bases in this manner avoids false impressions from an otherwise narrow sample group. Keep in mind that many users have more than one device...targeting one of those alone risks ignoring the others. As they say, limiting one's appreciation to only the most fragrant of flowers, severely hampers one's ability to spontaneously stop and smell the flowers.
As for which to buy, there is no simple fitzall. You haven't talked about what kinds of apps you intend to build. Just because you buy the single most popular device doesn't mean it will have the capabilities your apps claim to support. If you only intend to support iPhone, never iPad, then buy the newest feature laden phone you can afford and live large...at least until a newer model comes out. Keep in mind here that Apple still expects iPhone-only apps to work in 1x/2x emulation mode on iPad, tho. Get at least one of each and enjoy a yet again larger market.
I think in the end, (a) it is common among devs to have more than one test mule, and that's my hardware advice, and (b), as I mentioned, Apple expects devs to primarily focus on iOS version (software) over device model (hardware).
*If device diversity present$ a burden, you can always leverage TestFlight & users with devices you don't keep handy.