mulfib8 example circuit is underconstrained

[maltezellic]

The mulfib8 example is supposed to compute the n-th term of the multiplicative Fibonacci sequence, but the sequence is underconstrained for this purpose.
The transition constraints

result[0] = are_equal(next[0], current[6] * current[7]);
result[1] = are_equal(next[1], current[7] * next[0]);
result[2] = are_equal(next[2], next[0] * next[1]);
result[3] = are_equal(next[3], next[1] * next[2]);
result[4] = are_equal(next[4], next[2] * next[3]);
result[5] = are_equal(next[5], next[3] * next[4]);
result[6] = are_equal(next[6], next[4] * next[5]);
result[7] = are_equal(next[7], next[5] * next[6]);

only ensure that the next line is computed correctly from the last two entries in the current line.
However, the assertions

        let last_step = self.trace_length() - 1;
        vec![
            Assertion::single(0, 0, BaseElement::new(1)),
            Assertion::single(1, 0, BaseElement::new(2)),
            Assertion::single(6, last_step, self.result),
        ]

only fix the first two entries of the first row. The prover can thus set the rest of the first row to arbitrary values, in particular the last two entries, and thus change the result that is supposed to be the n-th term of the multiplicative Fibonacci sequence.

The fix should probably be to add constraints for the rest of the first row:

        let last_step = self.trace_length() - 1;
        vec![
            Assertion::single(0, 0, BaseElement::new(1)),
            Assertion::single(1, 0, BaseElement::new(2)),
            Assertion::single(2, 0, BaseElement::new(2)),
            Assertion::single(3, 0, BaseElement::new(4)),
            Assertion::single(4, 0, BaseElement::new(8)),
            Assertion::single(5, 0, BaseElement::new(32)),
            Assertion::single(6, 0, BaseElement::new(256)),
            Assertion::single(7, 0, BaseElement::new(8192)),
            Assertion::single(6, last_step, self.result),
        ]

All code is from examples/src/fibonacci/mulfib8/air.rs.