Don't mixup receiving and change addresses

[darosior]

[darosior]

I don't think we should have a separate derivation chain for change. Maybe just use odd addresses as change?

[darosior]

@kloaec made a good point about this: importing on another hypothetical wallet that would support our miniscript descriptor it might re-use a receive address as change.

I dislike to be forced into the boutique compatibility descriptor (with two derivation indexes in a single descriptor) and the legacy practice of using separate derivation paths just for supporting... Hypothetical address reuse on a future hypothetical wallet.

I don't know. "Be explicit about how to derive change/receive addresses in the descriptor" is a somewhat compelling argument.

[darosior]

I've opened rust-bitcoin/rust-miniscript#469 upstream to support multipath descriptors. I'm now leaning toward having a minimal compatibility layer here before this gets merged and released. But let's see.

[darosior]

So to upgrade to latest Miniscript we need another fix upstream. We need derived_descriptor to return a descriptor over a type that has the origin in addition to the raw pubkey. I'll do that.