FunC incorrectly optimizes expressions with operators: *, -, <=

[jeshecdom]

Based on tact-lang/tact#3019

Code to reproduce

int calc(int x) {
    return ((0 * (- x)) <= 0);
}

() main() {
    throw_if(16, calc(-115792089237316195423570985008687907853269984665640564039457584007913129639936));
}

Output Fift

// automatically generated from `func_test.fc` 
DECLPROC calc
DECLPROC main
calc PROC:<{
  DROP
  TRUE
}>
main PROC:<{
  256 PUSHNEGPOW2
  calc CALLDICT
  16 THROWIF
}>

Actual behavior when executed

calc always returns true.

Expected behavior when executed

Exit code 4, since - x should produce an overflow when x = -115792089237316195423570985008687907853269984665640564039457584007913129639936 (i.e., x = MIN_INT).

Side note 1: We can produce a related but different expression that also gets mapped into TRUE by replacing the subexpression - x for other expression that also produces overflows and the subexpression 0 for other expression that is equal to 0. For example, the following expressions also get transformed into TRUE in Fift, but they should produce error code 4 on the provided value for x:

• (~(-1) * (-1 * x)) <= 0 for x = MIN_INT.
• ((1 & (~ 1)) * (x / -1)) <= 0 for x = MIN_INT.
• ((x & 0) * (x * 2)) <= 0 for x = MAX_INT (i.e., x = 115792089237316195423570985008687907853269984665640564039457584007913129639935).
• ((x * 0) * (x + 1)) <= 0 for x = MAX_INT.

Side note 2: The related expression (0 * (- x)) == 0 DOES get translated to Fift in the expected way:

calc PROC:<{
  NEGATE
  0 MULCONST
  0 EQINT
}>