Racemisation of the reacting centre in the radical couplings

CHANGELOG.md

@@ -10,6 +10,7 @@ This follows the guideline on [keep a changelog](https://keepachangelog.com/en/1
 
 ### Fixed
 
+- `deoxygenative_coupling` and `cross_electrophile_coupling` no longer carry the reactant's configuration into the product at the reacting carbon. Both are Ni/photoredox couplings that break the bond homolytically and pass through a planar alkyl radical, so that centre is racemised — which is what makes the enantioconvergent variants of these reactions possible. The reacting carbon is now marked so RDKit discards its configuration: 120 of 120 `deoxygenative_coupling` templates at the carbinol carbon, and 52 of 196 `cross_electrophile_coupling` templates at whichever of the two coupling carbons is aliphatic (the remaining 144 couple two aromatic carbons and have no stereocentre to lose). Stereocentres elsewhere in the molecule are untouched, matching the chirality transfer reported for these reactions
 - Williamson products now have the correct configuration at the halide carbon. Williamson is an SN2 displacement of the halide, so that carbon inverts while the alkoxide carbon — whose C–O bond is never broken — is retained. Both were previously retained. All 104 williamson templates now mark the halide carbon (atom map 4) as inverted relative to the input; the alkoxide carbon is deliberately left unmarked. Verified against two reference outcomes: butan-1-ol + (S)-2-bromooctane gives (R)-2-butoxyoctane, and (S)-octan-2-ol + iodomethane gives (S)-2-methoxyoctane
 - Mitsunobu products now have the correct configuration at the carbinol carbon. The reaction proceeds by backside attack, so that centre inverts, but the templates left it unmarked and RDKit's default is to carry the reactant's configuration through unchanged — every stereodefined Mitsunobu product was the wrong enantiomer. All 60 mitsunobu templates now mark the centre as inverted relative to the input, so both input enantiomers are handled by the same template and an undefined centre stays undefined. No products are gained or lost; only their stereochemistry changes