Delete duplicate InMemoryPersister reimplementations across language bindings #1529

Each language binding reimplements InMemoryPersisters for the receiver and sender in tests, we could do a big code deletion there now that it's exposed.

Originally posted by @spacebear21 in #1528 (comment)

Fed this comment into my meta context for #1528 to produce this issue which should be a sufficient prompt for the run that closes it. I sanity checked the below prompt and find the direction, acceptable. Per-language commits were not my idea but sound sane and easier to review without conflating one problem for another.

In terms of sequencing this makes sense to do before payjoin-ffi-1.0, before someone hand-rolls another JsonSenderSessionpersister or if someone ~junior is exceedingly confident confident they can earn the glory of getting it merged it in one-shot review-then-merge cycle (feeling free to ask clarifying questions before requesting review).

PR #1528 promoted payjoin::persist::InMemoryPersister to public, unconditional Rust
API and re-exported it from payjoin-ffi (payjoin-ffi/src/lib.rs:14). However,
payjoin-ffi exposes session persistence to bindings via the
JsonReceiverSessionPersister / JsonSenderSessionPersister traits (defined in
payjoin-ffi/src/receive/mod.rs and payjoin-ffi/src/send/mod.rs with
#[uniffi::export(with_foreign)]). Bindings implement those traits in their own
language, and each one currently reimplements an in-memory event log:

• Dart — InMemoryReceiverPersister, InMemorySenderPersister in
  payjoin-ffi/dart/test/test_payjoin_integration_test.dart
• Python — InMemoryReceiverSessionEventLog, InMemorySenderPersister in
  payjoin-ffi/python/test/test_payjoin_integration_test.py
• C# — InMemoryReceiverPersister, InMemorySenderPersister in
  payjoin-ffi/csharp/IntegrationTests.cs

These are functionally identical: append events to a list, return the list on load,
set a closed flag.

Proposed work (stage)

1. payjoin-ffi — add uniffi-exposed in-memory persister types that implement
   JsonReceiverSessionPersister, JsonReceiverSessionPersisterAsync,
   JsonSenderSessionPersister, JsonSenderSessionPersisterAsync and wrap the Rust
   payjoin::persist::InMemoryPersister. Suggested names: InMemoryReceiverPersister,
   InMemoryReceiverPersisterAsync, InMemorySenderPersister,
   InMemorySenderPersisterAsync. Gate them behind the existing
   payjoin-ffi/_test-utils feature so they ship to binding test scripts but don't bloat
   the production FFI surface.

2. Per-language cleanup PRs (one each, mergeable independently after step 1):

   • Dart — delete the two test classes, regenerate uniffi bindings, swap call sites
   • Python — same
   • C# — same

Each binding PR is pure deletion + regen + import-swap once step 1 lands. Mechanical.

Why split

Step 1 needs core/FFI review attention (uniffi annotations, feature gating, async
variant). Steps 2a/b/c need per-language test verification (Dart hooks, Python uv +
bindgen, C# dotnet + bindgen). Splitting also lets the binding owners review their
own surface.

Out of scope

• Deduping anything else across bindings (this issue is scoped strictly to the
  InMemoryPersister reimplementations spacebear flagged)
• Production code changes — the in-memory variant is for tests/examples/prototyping only