Compliance wave 8: enroll conjoined event tenancy and subscriptions

[jeremydmiller]

Adopts JasperFx 2.45.0 (jasperfx#642) and enrolls the last two suites in the event sourcing compliance backlog. The shared library goes to 28 suites / 230 tests, both stores, still zero capability gates — and with these two, the ES backlog filed under #5118 is empty.

The two suites

ConjoinedEventTenancyCompliance (#5148) — the issue priced this as needing a tenant-scoped session seam. It didn't. OpenSession(IEventDatabase, tenantId) is already on the shared generic IEventStore<,>, and TenancyStyle is already a shared JasperFx.MultiTenancy enum. The only seam addition is the ComplianceStoreConfig.ConjoinedEventTenancy switch, which MartenComplianceFixture maps onto Events.TenancyStyle.

SubscriptionCompliance (#5151) — the one suite an empty subclass cannot close by itself. Both products declare ISubscription with an identical member, but ProcessEventsAsync returns a per-product IChangeListener, so that signature cannot be written once upstream. The shared ComplianceSubscription is therefore a partial class that each consumer completes; Marten's half is ComplianceSubscription.Marten.cs.

That file lives in Marten.Testing beside MartenComplianceFixture rather than in EventSourcingTests on purpose: both assemblies reference the source-only compliance package, so both compile the library's half of the partial, and both need the completing half to satisfy it.

Verification

Run against the published 2.45.0 package, not the -p:ComplianceSourceDir= dev loop.

• 230/230 compliance tests, zero skips (was 216 at wave 7)
• All 28 library suites confirmed enrolled — diffed the package's suite files against the subclass declarations rather than trusting the test count. A bump that satisfies a new registrar member is not evidence the suite runs: that is exactly how StrongTypedIdentityCompliance shipped 11 tests unrun in wave 6 (Enrol StrongTypedIdentityCompliance (#5144) #5198).
• Full EventSourcingTests: 1796 passed / 0 failed / 7 skipped on net9.0 — the 7 skips are the pre-existing baseline, unchanged.

Notes for the reviewer

• The library changes in 2.45.0 are compliance sources only, so the bump is behaviourally inert for the shipping Marten libraries. The full ES run above is the regression check on that claim.
• Polecat is a wave behind again (polecat#427 merged wave 7 earlier today, leaving it at 216/26). Its wave-8 catch-up is separate work in that repo.

Closes #5148
Closes #5151

🤖 Generated with Claude Code

https://claude.ai/code/session_01RrQZYcL13PEek7FErj9PGx