Integer // % to_base go through float64: silent wrong answers above 2^53 #149

From Motoko's #148 review. (// 123456789012345678 1000) returns ...344, (% MaxInt64 10) gives a false overflow, and MinInt64 // -1 gives CONVERSION_ERROR. Fix: real int64 math, or VALUE_ERROR for |a| >= 2^53. Also: let destructuring with a list of the wrong length silently fills nil or drops items; make it a VALUE_ERROR.