[cppyy] Don't leak deprecation warnings when resolving enum types

Cppyy::ResolveEnum() determines an enum's underlying integer type by
JIT-compiling a probe expression of the form:

```c++
std::is_same<unsigned int, std::underlying_type<SomeEnum>::type>::value;
```

If the enum being resolved (or its enclosing scope) is marked deprecated,
this purely internal introspection emits a `-Wdeprecated-declarations`
warning to stderr. Since cppyy resolves enums while wrapping classes, the
warning surfaces in ordinary PyROOT sessions that merely touch a class
carrying such an enum.

This started breaking a range of roottest tests (which diff process output
against reference files) after the historical ESTLType/ESTLtype enum copies
in TClassEdit, TDictionary and TStreamerElement were deprecated.

Wrap the probe expression in clang diagnostic push/ignored/pop pragmas so
that resolving a deprecated enum stays silent.

🤖 Done by AI.

Author: guitargeek

bindings/pyroot/cppyy/cppyy-backend/clingwrapper/src/clingwrapper.cxx

@@ -622,11 +622,17 @@ std::string Cppyy::ResolveEnum(const std::string& enum_type)
         std::ostringstream decl;
     // TODO: now presumed fixed with https://sft.its.cern.ch/jira/browse/ROOT-6988
         for (auto& itype : {"unsigned int"}) {
-            decl << "std::is_same<"
+        // This is pure type introspection: silence any deprecation warning that
+        // would otherwise be emitted just because the enum being resolved (or its
+        // scope) happens to be marked deprecated.
+            decl << "_Pragma(\"clang diagnostic push\")"
+                    "_Pragma(\"clang diagnostic ignored \\\"-Wdeprecated-declarations\\\"\")"
+                 << "std::is_same<"
                  << itype
                  << ", std::underlying_type<"
                  << et_short
-                 << ">::type>::value;";
+                 << ">::type>::value;"
+                    "_Pragma(\"clang diagnostic pop\")";
             if (gInterpreter->ProcessLine(decl.str().c_str())) {
             // TODO: "re-sugaring" like this is brittle, but the top
             // should be re-translated into AST-based code anyway