`DescriptorId` is not unique enough

[evanlinjin]

Describe the bug

As mentioned on Discord#development, there is a fundamental problem with how we define a DescriptorId. Currently DescriptorId does not uniquely identify a descriptor. I.e. we can change how the xpub is portrayed and it would completely change the DescriptorId. This is a problem because the whole point of DescriptorId is so that the last derivation index is stored with something that uniquely represents a descriptor.

Proposed Solutions

An idea (first mentioned to me by @LLFourn) is to use the hash of the first derived spk for the DescriptorId. We can consider disallowing non-wildcard descriptors in KeychainTxOutIndex (as they aren't keychains) and it allows us full uniqueness for DescriptorIds.

[notmandatory]

I like the proposal to use a hash of the descriptor's first derived SPK as the unique descriptor Id.

I don't know if anyone needs non-wildcard descriptors, but it feels like a regression to not allow them. What if we don't mix wildcard and non-wildcard keychains in the same KeychainTxOutIndex?

[notmandatory]

Since this is an edge case and doesn't affect the external API can I move this to the beta milestone? We can put a caveat in the beta release notes that this is a known issue. Or we can partially fix this by using the first SPK for each keychain, but document the open issue that using wildcard and non-wildcard descriptors that overlap are not supported.

[notmandatory]

@evanlinjin I tried implementing your proposed solution in #1486, let me know what you think. We shouldn't have to disallow non-wildcard xkeys as long as we clearly document that the user must ensure they have unique parent keys or derivation indexes other than the last index.