Don't attempt to parse dicts if not replacing

When reading in the config and processing dict content we were splitting
on = assuming we would always have key = value pairs. However in the
case where we don't want to replace because the testenv will not be used
this can result in ValueErrors as substitution variables are not string
key = value pairs.

Avoid all this by only attempting to process dict key = values if we are
replacing in the first place. Otherwise the content isn't needed and we
just return the default value or empty dict.

This should fix tox-dev#595

tox/config.py

@@ -957,16 +957,17 @@ def getlist(self, name, sep="\n"):
 
     def getdict(self, name, default=None, sep="\n", replace=True):
         value = self.getstring(name, None, replace=replace)
-        return self._getdict(value, default=default, sep=sep)
+        return self._getdict(value, default=default, sep=sep, replace=replace)
 
     def getdict_setenv(self, name, default=None, sep="\n", replace=True):
         value = self.getstring(name, None, replace=replace, crossonly=True)
-        definitions = self._getdict(value, default=default, sep=sep)
+        definitions = self._getdict(value, default=default, sep=sep,
+                                    replace=replace)
         self._setenv = SetenvDict(definitions, reader=self)
         return self._setenv
 
-    def _getdict(self, value, default, sep):
-        if value is None:
+    def _getdict(self, value, default, sep, replace=True):
+        if value is None or not replace:
             return default or {}
 
         d = {}