The outcome of Romance velar palatalization in French depends on the voicing of the consonant: Lat. ankilla → OFr. [antsele] but Lat. argilla → OFr. [ardʒile]. This is also reflected in words borrowed into English ("cent" with [s], "gentle" with [dʒ]). Is there any other language where historically or synchronically, the place of articulation of only one of /tʃ, dʒ/ changes. Framed more generally, where else does voicing affect place of articulation in alveopalatals? Either sound changes or synchronic rules are relevant.
I don't think it's at all common for alveolopalatal and/or sibilant consonants to undergo place changes that are directly conditioned by the presence or absence of voicing: French doesn't actually seem to be an example of this (see below) and I can't think of any actual examples, except for the disputed case of Proto-Semitic *z, which if pronounced [z] may have evolved differently from its voiced counterpart [s] in some Semitic languages.
I think that it's more likely for voice to act as the condition for a change in manner of articulation, which in turn conditions a change in place of articulation. French, Greg Lee's example of rhotacism, and amegnunsen's example of Zenete [ʃ] and [j]/[ʒ] all look compatible with "two-part" sound changes like that. ("Two-part" in theoretical, not necessarily chronological terms.)
Analyzing the French example
I have the impression that the difference between the French reflexes of Latin* kʲ and gʲ wasn't due to a change [t͡ʃ, d͡ʒ] > [t͡s, d͡ʒ], but rather because Latin kʲ developed a sibilant pronunciation earlier than gʲ: the data seems consistent with gʲ first passing through a stage where it was lenited to something like [j~ʝ] in all contexts, and then syllable-initial [j~ʝ] was fortited to [d͡ʒ], presumably after the voiceless sound had already been fronted to [t͡s].
That at least is what the Wikipedia article "Phonological History of French" says, although it doesn't have many citations.
So it might be something like [anˈkɪlla], [arˈgiːla] > [anˈcel-], [arˈɟi-] > [anˈcel-], [arˈʝi-] > [anˈt͡ʃel-], [arˈʝi-] > [anˈt͡sel-], [arˈʝi-] > [anˈt͡sel-], [arˈd͡ʒi-].
Greek dialect fronting of velars/palatals
TKR's answer mentioned Greek, which is a language that has shown various forms of palatalization, fronting or affrication.
There is some discussion of variations of the development of fronted consonants from historical velars in modern Greek varieties in "Velar Fronting in Modern Greek Dialects", by Io Manolessou and Nikolaos Pantelidis.
Table 1 shows some specific examples that indicate that fronting of ɣ to a sibilant is less common in Greek than fronting of k or g (g is historically from /nk/ or /ng/) to sibilant affricates. This seems like it might be relevant to the situation with French. The table also indicates a few Greek dialects that are supposed to have a sibilant affricate realization for the voiceless but not the voiced plosive, but I don't know the exact situation with these.
(Pronouncing "velars" as nonsibilant palatals before front vowels is extremely common: it's mandatory in standard Greek, and the paper says that exceptional dialectal forms are "rare").
Semitic [z] and [s]? [g] and [k]?
The history of fricatives in Semitic languages seems to be unclear and disputed, but there definitely have been some place changes. Depending on what viewpoint you take, voiced "z" may not have behaved the same as its voiceless counterpart. This seems like anything but a clear example, as the place of articulation of the Proto-Semitic voiceless sibilants is disputed, and there are also arguments about which, if any, of the sibilants were actually affricates.
Another odd thing about voicing and POA in Semitic is the unconditional fronting of *g but not *k in standard Arabic, although there are different developments of these sounds in Arabic dialects.
*In this answer, I'll use the symbols kʲ and gʲ to represent the sounds that in Latin were spelled as c or g before a front vowel. They aren't meant to be IPA transcriptions of the exact phonetic quality, which might have been something that would better be represented by some other symbols, e.g. [c] and [ɟ].
Not quite parallel but at least similarly asymmetrical: in Ancient Greek, *ky > [tt] / [ss] depending on dialect, while *gy > [zd] / [dz]. Presumably both went through an affricate stage of some kind.
I wonder if the French situation has to do with the lesser cross-linguistic frequency of [dz] vs. [ts].
Another Ancient Greek example is the treatment of Proto-Indo-European labiovelars before /i/. The voiceless labiovelar *kʷ became dental /t/, while the voiced *gʷ became labial /b/:
- *kʷis 'who' > τίς
- *gʷíh₃os 'life' > βίος
(Apparently the voiced aspirated labiovelar *gʷʰ also went to a labial, but this actually runs counter to the pattern above because voiced aspirates were devoiced in Greek preceding this change, so the development would have been *gʷʰ > *kʷʰ > *pʰ. But there's only one example of this so it isn't really clear.)
Based on inscriptional and dialectal evidence, this change went through an alveopalatal stage, so it's very close to your French example of /tʃ, dʒ/ ending up in different POAs based on voicing.
In Berber, it is supposed diachronically that some varieties, named Zenete, changed the sounds *k and *g respectively by [ʃ] and [j]/[ʒ] (see Kossmann 1999).
Palatalisation affects also laterals. In Riffian, *lː is become [ʤː]. In other Berber languages, it is the simple lateral that has evolved into [j]/[dz] (Kabyle) or [ʒ] (Middle Atlas/ Mauritania). This feature also is present in Romance languages.
It should notice also that, in synchrony, in Zenete languages, some geminated palatalised consonants do not stay palatale consonants, but are plosive. For example:
Rhotacization of z (not s) in Germanic, Latin (honos, honoris). This is pursuant to the Law of Similarity, according to which sounds which are most similar to their surrounding sounds are the first to be affected by a change. In this case, the change is from obstruent (z) to sonorant when the obstruent has the same voicing as surrounding sounds (vowels, which are both sonorant and voiced).