Le Pen might actually have an outside shot at winning Mayotte, which is the only Muslim majority department. Albeit with very unique circumstances.
Other than that, you are never going to know for sure because it's not legal to collect those sorts of statistics - perhaps assume based on the results of certain areas known to have large populations with immigrant backgrounds. But even then, RN do quite well in some of the pavillonaire suburbs on the North-Eastern edge Seine-Saint-Denis or in parts of the Val d'Oise, and those are still areas with pretty large Maghrebin origin populations.
Though Le Pen would be lucky to come with 25 points of Macron, the RN have been stagnating at best since 2017 - so who Muslims vote for is a pretty safe bet.
1. What are the unique circumstances of Mayotte?In a nutshell there is a huge level of illegal immigration from the neighbouring Comoros, which has been continuing to increase even since 2017. Mayotte is already deeply impoverished and with a very fragile economy itself, so the social impacts of that migration is pretty explosive and obviously pretty useful for an anti-immigration themed political party.
2. I know the Census is forbidden from collecting ethnoracial statistics in France, but aren't there exit polls with religious crosstabs? I know that exit polls can be trash with crosstabs, but still.
Well, France doesn't do exit polling like happens elsewhere - what you get when the polls close is a projection based on what has already been counted at a sample of polling stations before the final ones actually do close.
Investigating a bit further Ipsos did
a retrospective poll in 2017, where the best you'll find is Macron winning 72-28 amongst "other religion", which is probably mostly Muslims, but will also include Jews, Protestants and so on and you'd never be able to spin out exactly what is what - basically done so in order to not look like they are trying to collect ethnic statistics
As in, ethnic stats are illegal, but the rules are ambiguous enough that you can make assumptions based on say "people with an arab-muslim first name", you just can't make it too explicit that you are doing what you are doing. So stuff exists, but is always a bit imprecise and ambiguous at best.
3. I would assume that most RN voters in those Maghrebin-heavy suburbs are people of native French ancestry?
I would assume too, but we can't know for sure, and a lot of those places do actually have very mixed populations, including very assimilated people of North African background and very high levels of mixed marriages (something like 40-45% of "women with at least one North African born parent" get married to someone with no migration background - very contrary to the existing stereotypes*). So by all means there is in these sorts of places a immigrant descended population that is very assimilated and given to behaving similarly to the other lower-middle/upper-working class types who live in those sorts of areas.
* source Hervé Le Bras and Emmanuel Todd in "Le mystère français" before anyone attacks me
