BLD: Add clang -ffp-exception-behavior=strict also for _so

Apparently, there is a second `compiler_so` that is actually the
main compiler being used, and modifying only the first one is just
futile.

numpy/distutils/ccompiler.py

@@ -389,6 +389,7 @@ def CCompiler_customize_cmd(self, cmd, ignore=()):
         # clang defaults to a non-strict floating error point model.
         # Since NumPy and most Python libs give warnings for these, override:
         self.compiler.append('-ffp-exception-behavior=strict')
+        self.compiler_so.append('-ffp-exception-behavior=strict')
 
     def allow(attr):
         return getattr(cmd, attr, None) is not None and attr not in ignore