Migrate to new roc

[ageron]

This PR migrates the Roc track to the new zig-based Roc compiler and its new syntax.

At the time of writing:

• 92 exercises have been migrated and all their tests pass.
• 4 exercises have been migrated but compilation fails: alphametics, complex-numbers, protein-translation and strain.
• 14 exercises are partially migrated (i.e., only the tests):
  • 7 exercises shouldn't be too hard to finish migrating: forth, go-counting, grep, poker, rational-numbers, tournament, and zebra-puzzle.
  • 7 exercises depend on external libraries, so they will be impossible to migrate until these libraries are migrated. This includes anagram (unicode), gigasecond (isodate), meetup (isodate), micro-blog (unicode), rest-api (json), reverse-string (unicode) and sgf-parsing (roc-parser)

[Anton-4]

Constructors are redundant in Roc, is there a special reason why it was introduced here?

[ageron]

Oh? But a {real, image} is not a Complex unless Roc can tell that it's the type we expect, right? For example, the following code causes an error because the conjugate method actually returns a {real, image}, not a Complex.

Complex := { real : F64, imag : F64 }.{
	conjugate = |z| { real: z.real, imag: -z.imag }  # Error
}

So I need to add a type annotation to get what I want:

Complex := { real : F64, imag : F64 }.{
	conjugate : Complex -> Complex
	conjugate = |z| { real: z.real, imag: -z.imag }  # Now OK
}

Such type annotations are not always very ergonomic, for example:

main! = |_| {
	z1 : Complex
	z1 = { real: 1.23, imag: 4.56 }
	z2 : Complex
	z2 = { real: 7.89, imag: 1.23 }
	z3 = z1 + z2
	dbg z3
	Ok({})
}

versus:

main! = |_| {
	z1 = Complex.new({ real: 1.23, imag: 4.56 })
	z2 = Complex.new({ real: 7.89, imag: 1.23 })
	z3 = z1 + z2
	dbg z3
	Ok({})
}

I believe Jared has been working on a nicer syntax like Complex({real: 123, imag: 4.56}): it would definitely remove the need for contructors. I propose to remove new once this syntax is available.