The [docs](https://docs.python.org/3/library/functions.html#round)
specify that calling `round` with a negative precision removes
significant digits, so that `round(12345, -2) == 12300`. The
implementation was simply returning the original integer.
Additionally, `round(a, b)` is implemented as `(a / 10^b) * 10^b`, using
half-even rounding during the division.
This also fixes a bug in the log2() implementation, where very large
negative integer arguments would not raise an exception (and instead
would turn into NaNs).
