math/big: Support precision of -1 in Rat.FloatString

[Thomasdezeeuw]

The strconv.FormatFloat function accepts a special precision of -1 to mean to use the smallest number of digits necessary such that ParseFloat will return float exactly. This is a feature request to accept the same for math/big.Rat.FloatString.

[gopherbot]

Change https://golang.org/cl/344729 mentions this issue: math/big: Support precision of -1 in Rat.FloatString