I've heard all the voice actors, at least in vanilla, were Canadian.
It's my understanding that not all were, but they certainly gave priority to Canadian voice actors because they were recorded in Canada and that would save them a ton of money.
Specifically, Corsican French.
Yeah, it's worth pointing out that his French would have sounded heavily Italian. My personal view is that would have been terrible. He's in the game to give life to the French language. Even though the real Napoleon had a very different accent than most French people, that little detail needs to be ignored, imo.
Pretty sure we know what Gandhi and Haile Selassie sounded like.
That's true. From what I recall, Haile Selassie doesn't sound too bad. Interestingly, circling back to my previous point, Gandhi apparently spoke English almost exclusively as a modernizing thing. But, in the game, it's better for flavor for him to speak Hindi.
OP: Attila's language (Chuvash) contains some very broken grammar and it's apparent that the voice actor had no idea what any of the words meant.
Attila was who I had in mind when I said that the Civ5 language system wasn't perfect. To me, we know so little about the Huns that it probably wasn't a good idea to include them. I've heard (from probably biased Slavic sources) a good argument that the Huns were Slavs. There's the still held but widely discredited argument that they were the Xiongnu (pronounced "Hunnu" at the time). They went with a Steppe language that's as good a guess as any, but there are flaws in both the grammar and pronunciation, which suggests both the translator and voice actor (always separate people in Civ5) didn't quite know the Chuvash language.
Also worth noting that Harun, Ramesses, and al-Mansur are all speaking with different dialects. Harun is using modern peninsular Arabic, Ramesses is using Egyptian Arabic filtered through an amazing VA, and al-Mansur is using Moroccan Arabic.
Yeah, it's worth pointing out that, were it not for the Quran and shared common heritage, these Arabic dialects would almost certainly be considered separate languages. They're barely intelligible if you speak slowly. People who speak Arabic who tried to translate Al-Mansur's lines seemed to struggle more than someone who spoke Hebrew translating Dido's Phoenician lines.
However, if he's speaking Classical Arabic, that's different. I didn't recognize that it was the same language at first because he uses a lot of different words and says things very differently. I suspect, however, that they used classical Arabic to save money on the translator even if they got an Egyptian to say the lines. In that sense, Egyptian Arabic would have been better for greater diversity, but that's criticizing the solution they came up with for not being the second best solution while the heart of the problem is still that it's not Egyptian.