ICU-20575 fix broken default locale mapping for C.UTF-8

- Correct regression in  1afef30 / ICU-20187
- PR#418 incorrectly dropped the mapping from C to en-us-POSIX without handling
   cases where a codepage was set (such as C.UTF-8).
- The “strcmp("C", …)” in uprv_getPOSIXIDForCategory() was intended to detect
specific platform behavior, it is not a general mapping.
- Add an additional test in uprv_getDefaultLocaleID() for locales that end up
as "C" or "POSIX" after removing codepage suffix

Author: srl295

icu4c/source/common/putil.cpp

@@ -1560,6 +1560,10 @@ static const char *uprv_getPOSIXIDForCategory(int category)
     {
         /* Nothing worked.  Give it a nice POSIX default value. */
         posixID = "en_US_POSIX";
+        // Note: this test will not catch 'C.UTF-8',
+        // that will be handled in uprv_getDefaultLocaleID().
+        // Leave this mapping here for the uprv_getPOSIXIDForDefaultCodepage()
+        // callers.
     }
     return posixID;
 }
@@ -1681,6 +1685,11 @@ The leftmost codepage (.xxx) wins.
          * Many of the other locales using ISO codes will be handled by the
          * canonicalization functions in uloc_getDefault.
          */
+    } else if ((uprv_strcmp("C", correctedPOSIXLocale) == 0) // no @ variant
+               || (uprv_strcmp("POSIX", correctedPOSIXLocale) == 0)) {
+      // Raw input was C.* or POSIX.*, Give it a nice POSIX default value.
+      // (The "C"/"POSIX" case is handled in uprv_getPOSIXIDForCategory())
+      uprv_strcpy(correctedPOSIXLocale, "en_US_POSIX");
     }
 
     if (gCorrectedPOSIXLocale == nullptr) {