Yes, rhotic metathesis (or r-metathesis) is a real thing. If you want a recent-ish study on it, try this 2013 paper by Bartłomiej Czaplicki.
Simply, metathesis is the transpositions of letters, sounds, or syllables within a word. You see it everywhere in many languages worldwide, and rhotic metathesis is no exception. In his comment to your question, Gaston mentions a few of the more well-known examples in English, and you do frequently see the phenomenon going from Latin to Old French, such as:
turbulare -> turbler -> trubler (Eng. trouble)
fimbria -> frimbia -> frenge (Eng. fringe)
While this principle is sound, it's not quite what happened with the verbs in question. First, no one knows exactly where harpazein comes from, but the initial h-, if it's IE and related to harpe (see Beekes on why that might not be the case), it would likely begin with initial s-. But this isn't for sure.
Now, people used to think that harpazein is cognate with rapere, but that's no longer current. (LSJ includes it; Beekes mentions it at ἐρέπτομαι.) Regardless, both the Latin rapere and the German rauben (cognate with rob) do not directly descend from the Greek; they are derived more immediately from their parent languages. The Proto-Germanic *raubōną gives rise to the German rauben and the English rob. Both go back to PIE, with the initial letter being hr-.
If there is a Latin connection, Wiktionary points to *Hrewp- being cognate with the Latin rupes ("cliff") and rumpere ("to shatter") (with a present nasal infix; past passive participle ruptus).
Following Draconis' comment, we do see a cognate with rapere in Greek's ἐρέπτομαι. This is a normal yod-present (or jot-present) derivation of the PIE ancestor. It is not the direct ancestor of the Latin rapere or the German/English rauben/rob.
To sum, yes, rhotic metathesis is real and there is plenty of evidence for it. No, that's not what's going on with these verbs. Crucially, though, harpazein does not give rise to rapere and (either directly or indirectly) rauben and rob.