Co-authored-by: Cecile Tonglet <cecile.tonglet@cecton.com>

Assuming the `T2`-`T12` tuple `impl` enumeration matches [`impl
From<(...)> for [T;
...]`](https://doc.rust-lang.org/std/primitive.array.html), current code
misses not so obvious single-item tuple
[`(T1,)`](https://doc.rust-lang.org/std/primitive.array.html#impl-From%3C(T,)%3E-for-%5BT;+1%5D).

The style (trailing comma) of writing single-item tuple in the `impl` is
made to be the same as actual single-item tuple syntax (although did not
really have to be), macro associated with this also was changed, to
allow for single-item tuple (change turns `(A, B, C)` into `(A, B, C,)`,
which is the same type for all tuples except single-item`) and its
trailing comma syntax.

This PR implements `PartialOrd` and `Ord` for `IString`, fixes #18,
additionally adds more implementations for reverse comparisons (`str =
IString`, as opposed to usual `IString = str`).

It adds an `impl_cmp_as_str!` macro which allows to list implementations
easily by mentioning the trait name and pairs for which to implement the
trait.

It also updates tests using macros to list more possible usage variants
(comparing `IString`s of different `enum` variants for all tests,
reverse comparisons, adding tests for `PartialOrd` and `Ord`).

Clippy added a new lint `clippy::non_canonical_partial_ord_impl` (aka `incorrect_partial_ord_impl_on_ord_type`), hence it should now be followed.

This approach still allows using the same files for Rc and Arc while
preventing imports from leaking as well as working better with
rust-analyzer.