Fold accents to ASCII in the heading slug

A French deck anchored "Données" as "donnes" and "nœuds" as "nuds": the slug
kept only [a-z0-9_- ], so every accent was deleted rather than folded. NFD now
splits each letter from its accents and the marks are dropped, with the two
ligatures that have no canonical decomposition spelled out.

index.html and check_tutorial.py fold identically, diffed over accents, ligatures
and the punctuation that must still vanish. NFD, not NFKD: NFKD would rewrite
degree signs and ellipses and move anchors that already ship. Verified no
existing anchor moves — the 45 headings of the two English tutorials slug
byte-identically before and after.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: fabien-simvia

01_preprocessing/CAO_Maillage_Shaper_SMESH/README.md

@@ -59,7 +59,7 @@ The course runs in five parts, in the order the original deck presents them:
 
 > **In this part**
 > [Les grands principes](#les-grands-principes) ·
-> [Une CAO pour la simulation numérique](#une-cao-pour-la-simulation-numrique) ·
+> [Une CAO pour la simulation numérique](#une-cao-pour-la-simulation-numerique) ·
 > [Redessiner une CAO existante](#redessiner-une-cao-existante) ·
 > [Modifier une CAO existante](#modifier-une-cao-existante) ·
 > [Les tests](#les-tests) ·
@@ -300,11 +300,11 @@ The course runs in five parts, in the order the original deck presents them:
 
 > **In this part**
 > [Objectif](#objectif) ·
-> [Données](#donnes) ·
+> [Données](#donnees) ·
 > [L'esquisse de la plaque](#lesquisse-de-la-plaque) ·
 > [Les contraintes](#les-contraintes) ·
 > [La face et les groupes](#la-face-et-les-groupes) ·
-> [Le maillage : algorithme et hypothèses](#le-maillage--algorithme-et-hypothses) ·
+> [Le maillage : algorithme et hypothèses](#le-maillage--algorithme-et-hypotheses) ·
 > [Calcul et exportation](#calcul-et-exportation)
 >
 > *Original deck: slides 14–42.*
@@ -896,9 +896,9 @@ The course runs in five parts, in the order the original deck presents them:
 ## Part 5 · Exercice n°2
 
 > **In this part**
-> [Objectif du tube coudé](#objectif-du-tube-coud) ·
-> [Données du tube coudé](#donnes-du-tube-coud) ·
-> [Principe général](#principe-gnral) ·
+> [Objectif du tube coudé](#objectif-du-tube-coude) ·
+> [Données du tube coudé](#donnees-du-tube-coude) ·
+> [Principe général](#principe-general) ·
 > [Analyse](#analyse) ·
 > [Les esquisses du tuyau](#les-esquisses-du-tuyau) ·
 > [Le chemin et le tuyau](#le-chemin-et-le-tuyau) ·

index.html

@@ -155,8 +155,16 @@ <h2>On this page</h2>
 const bar = document.querySelector("#bar i");
 
 // GitHub-style heading anchors, so the in-page links in the READMEs work.
+// Accents are folded to ASCII before the class below drops everything outside
+// [a-z0-9_- ]: a French deck would otherwise anchor "Données" as "donnes" and
+// "nœuds" as "nuds". NFD splits a letter from its accents, \p{M} deletes the
+// accents, and the two ligatures that have no decomposition are spelled out.
+// check_tutorial.py must fold identically — see the note on slug() there.
 const slug = (s) => s.toLowerCase().trim()
-  .replace(/<[^>]+>/g, "").replace(/[^\w\- ]/g, "").replace(/ /g, "-");
+  .replace(/<[^>]+>/g, "")
+  .replace(/œ/g, "oe").replace(/æ/g, "ae")
+  .normalize("NFD").replace(/\p{M}/gu, "")
+  .replace(/[^\w\- ]/g, "").replace(/ /g, "-");
 
 function render(md, base) {
   const code = [], math = [];