Consider this snippet when compiled as C++:

void test(int n, int m) {
  int x = n < m;
  if(x) {
    
  } else {
    
  }
}

In C++ code we have the following conversions:

1. A bool-to-int conversion on the result of n < m
2. An int-to-bool conversion on x in if(x)

We handled the bool-to-int conversion (i.e., the first conversion) perfectly fine. But the second conversion means that we're not branching on x, but rather on x != 0. And we didn't have any guard logic to handle the fact that x != 0 implied n < m.

This PR adds that logic.

Commit-by-commit review recommended:

• The first commit improves the tests a bit
• The second commit adds some more tests
• The third commit is really the meat of this PR
• The fourth commit simply accepts test changes

(Eventually I would like to move C/C++ over to the new shared guards library, but since that's a much larger task I would like to delay that.)