Add test checking if trace bug in field sensitivity was fixed

The test in this commit checks for a bug in which the initial nondet
assignment to struct fields was missing from --trace (always showing
these struct fields as being initialised to 0).

Author: hannes-steffenhagen-diffblue

regression/cbmc/field-sensitivity-trace-wrong-counterexample-1/test.c

@@ -0,0 +1,14 @@
+struct SomeStruct
+{
+  int x;
+};
+
+int main(void)
+{
+  struct SomeStruct x;
+  // this should cause
+  // visible initialization of
+  // x to value other than 0
+  // if assertion fails
+  assert(x.x == 0);
+}

regression/cbmc/field-sensitivity-trace-wrong-counterexample-1/test.desc

@@ -0,0 +1,13 @@
+KNOWNBUG
+test.c
+--trace
+^EXIT=10$
+^SIGNAL=0$
+^VERIFICATION FAILED$
+x=\{ \.x=-?[123456789]\d* \}
+--
+^warning: ignoring
+--
+This tests for the counterexample shown in the trace having sensible values.
+In this case, we are asserting that x.x == 0, so x.x should have a value other
+than 0 if the assertion fails.