Ruby opaque pointer handling regression fix

This bug was introduced in swig-3.0.8 in #146252 adding shared_ptr
support. An ObjectPreviouslyDeleted error was incorrectly thrown
when the pointer was used as a parameter after being set to zero
via a call to 'DATA_PTR(self) = 0'.

It isn't clear to me which approach is better in this corner case,
so I've gone for backwards compatibility and restored the old behaviour.

Closes #602

Examples/test-suite/ruby_manual_proxy.i

@@ -0,0 +1,66 @@
+%module ruby_manual_proxy
+
+
+%typemap(in, numinputs=0) SWIGTYPE ** ($*1_ltype temp) "$1 = &temp;";
+
+%typemap(argout) SWIGTYPE **OUTPARAM {
+  $result = SWIG_Ruby_AppendOutput($result, SWIG_NewPointerObj(*$1, $*1_descriptor, 0));
+}
+
+%apply SWIGTYPE **OUTPARAM {
+  svn_fs_t **
+};
+
+%typemap(check) svn_fs_t * {
+  if (!$1) {
+    svn_swig_rb_raise_svn_fs_already_close();
+  }
+}
+
+%{
+typedef struct svn_fs_t {
+  char path[256];
+} svn_fs_t;
+
+void svn_fs_create(svn_fs_t **fs_p, const char *path) {
+  svn_fs_t *fs = (svn_fs_t *)malloc(sizeof(svn_fs_t));
+  strncpy(fs->path, path, 256);
+  *fs_p = fs;
+}
+const char *svn_fs_path(svn_fs_t *fs) {
+  return fs->path;
+}
+%}
+
+typedef struct svn_fs_t svn_fs_t;
+void svn_fs_create(svn_fs_t **fs_p, const char *path);
+const char *svn_fs_path(svn_fs_t *fs);
+
+%{
+static void svn_swig_rb_raise_svn_fs_already_close(void) {
+  rb_raise(rb_eIOError, "already closed");
+}
+
+static VALUE svn_fs_swig_rb_close(VALUE self) {
+  if (!DATA_PTR(self)) {
+    svn_swig_rb_raise_svn_fs_already_close();
+  }
+
+  DATA_PTR(self) = NULL;
+
+  return Qnil;
+}
+
+static VALUE svn_fs_swig_rb_closed(VALUE self) {
+  return DATA_PTR(self) ? Qfalse : Qtrue;
+}
+%}
+
+%insert("init") %{
+  {
+    VALUE cSvnfs;
+    cSvnfs = rb_const_get(_mSWIG, rb_intern("TYPE_p_svn_fs_t"));
+    rb_define_method(cSvnfs, "close",
+                     VALUEFUNC(svn_fs_swig_rb_close), 0);
+  }
+%}