add "long double" overloads of min/max for MSVC to avoid choosing

the "long" overloads for floating-point numbers


[SVN r13041]

libs/config

@@ -1 +1 @@
-Subproject commit 69ad1282a3302ab91480273341e75270ec4e0f96
+Subproject commit 7e7c15a4cc052e2f193f75bf7123a942a33aa0d8