Persist cached SPKs on wallet creation

[j-kon]

Summary

• flush the indexer's staged SPK cache during wallet creation so persisted wallets write the initial lookahead cache immediately
• avoid emitting empty staged cache entries when reloading wallets with persisted SPK cache enabled
• add a regression test proving a wallet created with use_spk_cache(true) does not need an initial reveal before the cache persists

Closes #387.

Testing

• cargo test --all-features --test persisted_wallet
• cargo test --all-features

cc @ValuedMammal, @oleonardolima

[ValuedMammal]

The main issue is with the spk_cache_stage AFAICT, because it alters the ChangeSet semantics in subtle and surprising ways. It's not clear to the developer why indexing a null txout would be necessary to "flush the stage".

The SPK cache already acts as a kind of stage. What we want is to pass a &mut ChangeSet directly to replenish_inner_index and then ideally return it from insert_descriptor (or similar). At the very least we can just stage the indexer initial_changeset when creating a Wallet. Then reveal_next_address will only persist 1 newly derived SPK at a time.

Second, make_indexed_graph appears to be doing double duty by constructing the IndexedTxGraph from a changeset during both the creation and loading phases. This is unnecessary since create_with_params just passes an empty changeset. Further, we already assert that the stage must be empty upon loading, so what is the intention of the &mut stage in make_indexed_graph 🤔

[j-kon]

Thanks, this makes sense.

I reworked the patch to follow that direction and pushed an update. Instead of forcing the staged SPK cache out via index_txout, wallet creation now stages index.initial_changeset() directly, so the initial lookahead cache is persisted explicitly at creation time.

I also split the creation and loading paths: creation now builds the index directly and stages its initial changeset, while loading reconstructs the IndexedTxGraph from the persisted changeset without producing new staged changes. That removes the &mut stage coupling from the load path and keeps the "stage is empty upon load" invariant intact.

The regression test is still covering the original issue: a wallet created with use_spk_cache(true) can be reloaded immediately, and the first reveal_next_address only stages the newly derived SPK rather than restaging the initial cache.

I reran cargo test --all-features --test persisted_wallet and cargo test --all-features after the refactor.

[codecov]

Codecov Report

❌ Patch coverage is 97.67442% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 80.06%. Comparing base (4e202c8) to head (9d78337).
⚠️ Report is 16 commits behind head on master.

Files with missing lines	Patch %	Lines
src/wallet/mod.rs	97.67%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #428      +/-   ##
==========================================
+ Coverage   80.04%   80.06%   +0.02%     
==========================================
  Files          24       24              
  Lines        5336     5348      +12     
  Branches      242      242              
==========================================
+ Hits         4271     4282      +11     
- Misses        987      988       +1     
  Partials       78       78              

Flag	Coverage Δ
rust	80.06% <97.67%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.