QCheck2.Gen: enforce naming consistency for type int

[vch9]

First PR of attempting to tackle https://github.com/c-cube/qcheck/pull/223 in multiple smaller mrs. It should facilitate reviews.

Notice that I move code around so int t generators are close to each other both in the implementation and interface.

Naturals

Generators becomes:

val nat : int t
val nat_small : int t
val nat_big : int t
val nat_corners : unit -> int t

Deprecated natural generators are:

val small_nat : int t
(** @deprecated use {!nat_small} *)

val big_nat : int t
(** @deprecated use {!nat_big} *)

Classic generators

Generators becomes:

val int : int t
val int_neg : int t
val int_pos : ?origin : int -> int t
val int_small : int t
val int_big : int t
val int_corners : unit -> int t
val int_pos_bound : int -> int t
val int_range : ?origin:int -> int -> int -> int t
val (--) : int -> int -> int t

Deprecated generators are:

val neg_int : int t
(** @deprecated use {!int_neg} *)

val small_signed_int : int t
(** @deprecated use {!int_small} *)

val small_int_corners : unit -> int t
(** @deprecated use {!nat_corners} *)

val pint : ?origin : int -> int t
(** @deprecated use {!int_pos} *)

val int_bound : int -> int t
(** @deprecated use {!int_pos_bound} *)

I then ran

git grep <generator> src/core/QCheck2.* test/core/QCheck2*

On each deprecated generator to see if all occurrences (expected the deprecation cycle) were removed.

TODO

• [ ] int_pos has ?origin but int_neg does not, makes it weird.
• [ ] CHANGELOG
• [ ] Potentially add regression tests for every uncovered tests

Question

Now that we have this smaller PR, we can question the existence of nat and/or int_pos @jmid you said in the last PR

I noticed that

• nat is non-uniform and at most 10.000 - while int is uniform over all OCaml ints. (one should use nat_origin 0 to get a behavior mirroring int)
• there's no int_big

I think we could change nat* to become kind of an alias to int_pos*. That'd mean that int_pos would be removed and underlying generators for naturals will become uniform over OCaml positive integers rather than "non-uniform and at most 10.000"