Auto merge of #147502 - camsteffen:split-overlapping-impls, r=fmease

Split overlapping_{inherent,trait}_impls

This yielded some perf improvement for me. Reduces some calls to `impl_trait_header` query. But I think the llvm optimization is more relevant.

compiler/rustc_hir_analysis/src/coherence/inherent_impls_overlap.rs

@@ -154,7 +154,7 @@ impl<'tcx> InherentOverlapChecker<'tcx> {
         impl1_def_id: DefId,
         impl2_def_id: DefId,
     ) -> Result<(), ErrorGuaranteed> {
-        let maybe_overlap = traits::overlapping_impls(
+        let maybe_overlap = traits::overlapping_inherent_impls(
             self.tcx,
             impl1_def_id,
             impl2_def_id,