test(db): use canonical channel roster fixtures

[TheSentinel454]

Summary

Fix two PostgreSQL channel-roster test fixtures so they satisfy migration 0032's canonical kind-39002 fence before the channel/membership store extraction moves them.

This is a test-only prerequisite in the issue #2 extraction stack. It emits the canonical four-field p tags with authoritative roles, keeps the stale-history mutation after insertion, and gives the colliding other-tenant fixture its own matching roster.

Stack

• Exact base: codex/issue-6-finish-replaceable-store at da01840 (#6777)
• Exact head: codex/issue-7-roster-test-fixtures at 21d1b26
• Structural tracker: buzz-db: break up lib.rs — move impl Db methods to domain modules TheSentinel454/buzz#2
• Domain issue: buzz-db: split channel.rs into channel CRUD + membership TheSentinel454/buzz#7
• Acceptance: buzz-db: move lib.rs tests (~4,000 lines) to domain modules TheSentinel454/buzz#17 and buzz-db: consistent #[datastore_span] tracing placement TheSentinel454/buzz#19
• Next: codex/issue-7-channel-membership-store

Why separate

The existing ignored tests are red against a freshly migrated current main: migration 0032 rejects their legacy two-field roster tags before the assertions run. The extraction PR is intended to be a pure move, so this fixture correction is isolated here rather than mixed into the channel-membership ownership diff.

Non-goals

• No production Rust or SQL changes.
• No schema, lock, transaction, timeout, retry, API, or client-visible behavior changes.
• No change to migration 0032's canonical roster rules.
• No store ownership movement; that remains in the child PR.

Risk

Low and test-only. The main risk is accidentally changing the scenario rather than only its representation. The tests retain the same member counts, stale-versus-complete distinction, tenant collision, signer isolation, and lock-freshness assertions.

Blox verification

Author workstation: buzz-tornquist-issue-2-store-stack (2046520), native PostgreSQL 17.11.

• Red before the change on a freshly migrated database: both focused tests failed with migration 0032's kind 39002 roster contains an invalid p tag constraint.
• cargo fmt --all --check: pass
• git diff --check: pass
• cargo clippy -p buzz-db --all-targets -- -D warnings: pass
• channel::tests::large_roster_reconciliation_candidates_respect_snapshot_count_and_signer: pass
• channel::tests::locked_member_snapshot_blocks_post_capture_membership_mutation: pass
• Cumulative exact-tip PostgreSQL matrix: pass

Superseded pre-comment restack verification

PR #6700 merged before publication completed. This layer was restacked onto current main through the exact parent named above; the final cumulative tip is 2ddcc8a. Cumulative author gates passed: formatting and diff checks; buzz-db and buzz-relay all-target clippy with -D warnings; DB lib 111 passed / 200 ignored; ownership 22/22; observability 1/1; the full isolated PostgreSQL domain matrix; and relay lib 910 passed / 49 ignored.

• Workstation: buzz-tornquist-pr-6819-final-review (2057618), fresh shallow checkout
• Base: ffbeaaf00810aa359ab85818ff4820f92263e45f
• Head: c60e793eadde79d9eab9f48bbb2ede0ad4831f9b
• Findings: none

Reviewed the test-only roster-fixture correction. The affected large-roster and owner fixture events now use canonical four-field p tags and preserve the intended roster cardinalities. The distinct other-community fixture remains isolated and the production implementation, SQL, spans, and API are untouched.

Verification: format and diff checks passed; buzz-db --all-targets clippy passed with -D warnings; DB lib tests passed (111 passed, 200 PostgreSQL tests ignored); ownership (1/1) and observability (1/1) guards passed; all 18 channel PostgreSQL tests passed on native PostgreSQL 17 with migrations 1-32 successful. Final worktree was detached at the exact head and clean.

Complete evidence archive SHA-256: da8379f4f0ae3989eb3573162f0f19cbde733400000756ab0f63af3322244d4b.

Comment-addressed restack

Review follow-up on #6777 removed only the low-value replaceable ownership source test. This PR was restacked onto its rewritten parent; its production patch is unchanged.

• Exact base: da018405cc83605362125c2d5e5a3f91492431ac
• Exact head: 21d1b265c133292e6707e766cd4204e6a43f08af
• Final cumulative tip: 6fa2f104d42c6ba85bdf62e7ccb74ceaf4a84f67
• Per-layer patch-ID and tree audits confirm this PR’s production diff is unchanged from its pre-comment head.
• Cumulative Blox gate: formatting and diff checks; strict buzz-db/buzz-relay Clippy; DB lib 111 passed / 200 ignored; ownership 21/21; observability 1/1; every moved PostgreSQL test; relay lib 910 passed / 49 ignored.
• Independent re-review at this exact head: no findings; fresh exact-parent/head Blox review passed fmt/diff, strict buzz-db Clippy, DB lib 111 passed / 200 ignored, observability, and all 18 channel PostgreSQL tests; relay compilation was not applicable to this test-only layer.

[wpfleger96]

🤖 Combined review from three independent passes (two source reviews + one E2E test run at exact head 5df440c): no findings.

• Test-only diff; all three touched fixtures now emit the canonical four-field p tags migration 0032 requires, with the stale snapshot correctly modeled via post-insert UPDATE (bypassing the INSERT-only fence, matching pre-fence data).
• The tenant-collision fixture fix is a strengthening: the other tenant now gets a roster matching its own membership instead of reusing the primary tenant's owner tag.
• Scenario semantics preserved (stale=1,000 / complete=1,501, signer isolation, decoy pagination pressure).
• E2E: both focused tests plus the full 18-test channel::tests:: Postgres group passed on a fresh fully-migrated database.