fix: prevent a potential buffer overflow

When the total number of read octets is greater than the max return
buffer size in bytes, we return a failure status (-1) instead of jumping
out of the loop. Otherwise, a number larger than the maximum return
buffer size might be returned upon success, which if unchecked by the
caller, might cause a buffer overflow.

Note that the call to `reader.read(&mut ret_buffer[total_bytes_read..])`
might return a value less than or equal to the available space in
ret_buffer, which would make total_bytes_read no larger than the max
return buffer size. However, we'd like not to make the code safer by not
depending on an assumption on the implementation of the `read` method.

src/lib.rs

@@ -99,8 +99,9 @@ pub extern "C" fn cureq_call(method: *const c_char, url: *const c_char, headers:
             break;
         }
         total_bytes_read += bytes_read;
-        if total_bytes_read >= max_ret_buffer as usize {
-            break;
+        if total_bytes_read > max_ret_buffer as usize {
+            // response is too large for ret_buffer
+            return -1;
         }
     };