Please feel free to comment and critique anything.

There has been a discussion in issue #555 about the return type of measure(). To keep the discussion focused and have concrete code to talk about I decided to submit this PR now but I'm really open to changes.

This translation uses tuple (possible: bool, moves: int, goalBucket: string, otherBucket: int). For tasks that are possible to solve the function is expected to return a tuple where possible is set to true, for impossible tasks the function should returns a tuple where the member possible is set to false (the unmodified default result will do) and the values of the other members are ignored by the tests.

I looked at the other tracks and how they do it (as far as I understand these languages):

Three of them cop out: The Java version, F# version, and Ruby version did omit the tests for the impossible tasks completely.

Two tracks use some sort of compound type with a flag member that indicates impossible tasks: The C version and Go version.
They are similar to my translation (in its current state).

A lot of them use exceptions: The Python version, C# version, Crystal version, JavaScript version, PHP version, PowerShell version, TypeScript version, Vim Script version, and VisualBasic version use exceptions for impossible tasks.
The Lua version wants a failed assert and the Wren version wants some sort of "abort". I don't know the two languages but I think these are essentially exceptions.

Three use some sort of "Optional": The Common Lisp version and Elm version require the function to return NIL/Nothing for impossible tasks.
The Rust version uses an Option that is None for impossible tasks.

One uses some sort of "Variant": The jq version requires two different types for possible and impossible tasks.

I tried to implement this exercise with an Object Variant but that got more complicated, I had to use a ref type and I couldn't figure out how to write simple tests.
But probably you folks know how to do that better, right @ynfle?

I also tried Option.
That is pretty close to this first version with the .isPossible member. I'm not sure which of the two I like more.
From an educational standpoint it would probably be nice to have an exercise with Option.