Deriver: qcheck and qcheck2

[vch9]

Closes#190

This PR introduces two derivation plugins:

• qcheck
• qcheck2

The code and test is duplicated, in the vast majority it's only a renaming to QCheck to QCheck2 (excepts for fun_nary)

- utop # type t =intvalgen : tQCheck.Gen.t=<fun>valarb : tQCheck.arbitrary=
{QCheck.gen =<fun>; print =None; small =None; shrink =None;
collect =None; stats =[]}
─( 15:46:16 )─<command2>─────
-utop # type t =int [@@deriving qcheck2];;
typet = intvalgen : tQCheck2.Gen.t=<abstr>

This it not ideal as we only derive generators then call make on them, the best would be to use arbitrary for primitive types as they also have printers, shrinkers etc.

However after #195, #208 and this one, I'm getting tired of writing boilerplate code, but I will eventually do this in the future :)

[tmcgilchrist]

Would introducing a VERSION module and changing these to use a to_module function improve this duplication?

module type VERSION = sig
type t
val version : t
val to_module : t -> label
end
type version = [`QCheck | `QCheck2]
let to_module = function
| `QCheck -> "QCheck"
| `QCheck2 -> "QCheck2"

[vch9]

Yes you're right, I believe we would never have more than 2 versions but that'd simplify the code a bit.

[vch9]

Although, I don't see the real benefit of the module type VERSION.
Isn't:

typeversion = [`QCheck | `QCheck2]
letto_module : version -> string =function|`QCheck -> "QCheck"|`QCheck2 -> "QCheck2"lettyversion=Ldot (Ldot (Lident (to_module version), "Gen"), "t")
...
moduleMake (Version : sig val version : version end) =struct

enough?

[tmcgilchrist]

It would be enough, I would go with that.
Making it a module type might be slightly tidier if more than 2 versions were expected or there were other places that needed that type. In my hacking it didn't seem worthwhile.

[vch9]

Used @tmcgilchrist suggestion to reduce the code duplication and rebased to master. The commits should be squashed on merge, I'm not sure each of them compile and it's not that important if they're split.