Inline descriptor inner types

[tcharding]

Patch 1 is trivial cleanup

From commit log of patch 2

The `ShInner` and `WshInner` enum's provide no benefit that I can
discern. Both include only tuple variants with private tuple fields. We
can therefore inline the enums into `Sh` and `Wsh` respectively with no
loss of type safety. Doing so marginally simplifies the code, the
biggest benefit is it stops devs wondering why there is an inner type.

[apoelstra]

ACK bbf5ee9

[apoelstra]

This is a breaking change but hopefully not too disruptive of one. There isn't really any way to do a deprecation stage.

Will let @sanket1729 take a look before merging.

[tcharding]

This is a breaking change but hopefully not too disruptive of one.

On that topic, whats your view on the status of miniscript, you said in the past that the 1.0 release was a mistake, do you consider the crate is in the same state as rust-bitcoin i.e., we can shake things up a bunch with the aim of trying to stabilise so we can stop making so many changes to the public API?

[apoelstra]

@tcharding yeah. I think a "real" 1.0 should even be delayed until we get rid of all the recursion, make it impossible to construct invalid objects, and maybe improve the in-memory representation of things.

[tcharding]

Cool cheers, plenty to do then.

[apoelstra]

ACK 21e3cbc

[tcharding]

Rebase. Also fix a couple of match statements that used match &self to use match *self as is customary.

[sanket1729]

@tcharding, this was done on purpose to prohibit invalid use of constructors. If these are directly used in enums, user can directly create Wsh::Ms(invalid minisript). An invalid miniscript could be something that is not does have a root level type B among other things.

The purpose of shifting to inners was to guard access strictly via constructors.

[tcharding]

Ahhh, so my commit log shows the problem

The ShInner and WshInner enum's provide no benefit that I can discern.

The problem is with me, there is a difference, as you state - maintaining the invariant. My bad.

Closing, thanks.

[tcharding]

Ahhh I feel kind of vindicated, it is not immediately apparent that this function maintains the [undocumented] invariants.

    /// Create a new p2sh descriptor with the raw miniscript
    pub fn new(ms: Miniscript<Pk, Legacy>) -> Result<Self, Error> {
        // do the top-level checks
        Legacy::top_level_checks(&ms)?;
        Ok(Self { inner: ShInner::Ms(ms) })
    }