[analyzer][NFC] Add ArrayBound tests to document casting bug

Add a few security.ArrayBound testcases that document the false
positives caused the fact that the analyzer doesn't model a cast from
`signed char` to `unsigned char`.

Author: NagyDonat

clang/test/Analysis/out-of-bounds.c

@@ -194,3 +194,34 @@ char test_comparison_with_extent_symbol(struct incomplete *p) {
   return ((char *)p)[-1]; // no-warning
 }
 
+int table[256], small_table[128];
+int test_cast_to_unsigned(signed char x) {
+  unsigned char y = x;
+  if (x >= 0)
+    return x;
+  // FIXME: Here the analyzer ignores the signed -> unsigned cast, and manages to
+  // load a negative value from an unsigned variable. This causes an underflow
+  // report, which is an ugly false positive.
+  // The underlying issue is tracked by Github ticket #39492.
+  return table[y]; // expected-warning {{Out of bound access to memory preceding}}
+}
+
+int test_cast_to_unsigned_overflow(signed char x) {
+  unsigned char y = x;
+  if (x >= 0)
+    return x;
+  // A variant of 'test_cast_to_unsigned' where the correct behavior would be
+  // an overflow report (because the negative values are cast to `unsigned
+  // char` values that are too large).
+  // FIXME: See comment in 'test_cast_to_unsigned'.
+  return small_table[y]; // expected-warning {{Out of bound access to memory preceding}}
+}
+
+int test_negative_offset_with_unsigned_idx(void) {
+  // An example where the subscript operator uses an unsigned index, but the
+  // underflow report is still justified. (We should try to keep this if we
+  // silence false positives like the one in 'test_cast_to_unsigned'.)
+  int *p = table - 10;
+  unsigned idx = 2u;
+  return p[idx]; // expected-warning {{Out of bound access to memory preceding}}
+}