As I understand it, babies are initially able to produce and hear differences in basically any sound, but will in due time come to hone in on whichever sounds are used in the language they're being raised with. Listening for only a few sounds and remembering how to produce only a few sounds reduces the cognitive load, basically.

I'm able to make a pretty broad range of sounds, but I'm not sure how exactly I gained this ability. A part of it was certainly taking the time to learn the International Phonetic Alphabet and actually consciously learning the mechanics of making different sounds, but another part of it I think was just always enjoying making weird mouth noises, beatboxing, mimicking things, or doing silly voices or accents. Growing up with two languages probably also helped, but I'm told I have a "slight foreign accent" in Norwegian, so the exact extent of that help is a bit questionable.

Anybody, in any case, is able to learn to make new sounds and distinctions. It's probably easier than you might think, although it also might take some perseverance for some people, and you might never get perfect native-like pronunciation — but why should you want that, anyways? Own your accent, I say.

I think the problem with learning new sounds is oftentimes just having bad teachers, though. For instance rolled R's actually are fairly common (non-phonemically) in American English, especially in "what'd" or sometimes other contractions ending in -t'd: /wətəd/ → /wəɾəd/ → /wəɾəɾ‿/ → /wəɾɾ‿/ → /wər‿/ — this coalescence of /ɾəɾ/ to /r/ is considered to be more widespread in African American Vernacular English compared to other forms of American English, though. Like if you've seen that viral video "I've Never Seen Cops Run This Fast" you probably noticed how the cameraman very prominently says "outta there" as "ou[r]ere" and "speed it up" as "spee[r]up".