fix(relay): tombstone kind:30620 event row on NIP-09 a-tag workflow deletion

[s00ly]

Problem

handle_a_tag_deletion's KIND_WORKFLOW_DEF branch deletes the workflows
projection row but never tombstones the underlying kind:30620 event row, so
REQ subscribers (e.g. buzz workflows list) keep receiving the deleted
definition. Fixes #717 — same bug shape as #714, scoped to the workflow kind
(the generic NIP-33 arm added for #714 intentionally excluded it).

Implementation

Tombstone the event row via soft_delete_by_coordinate on both deletion paths:

• UUID path: the a-tag d-value is the workflow id, which is also the event
  row's d-tag (handle_workflow_def requires d = workflow UUID).
• Name path: the a-tag d-value is a name, not the event coordinate, so we
  resolve the workflow row first and tombstone by its id.
• The UUID-path tombstone is unconditional, which also heals pre-fix rows whose
  projection is already gone but whose event row stayed live (re-sending the
  deletion now converges them).

The coordinate uses the a-tag's pubkey — the NIP-33 author — matching what
validate_standard_deletion_event authorized.

Verified nothing operational relies on the row staying live: the workflow
engine and the webhook bridge load definitions from the workflows table
(never the events row), and audit lookups go through
get_event_by_id_including_deleted, which is tombstone-tolerant by design.

Composes with #2242 (same function, different concern: that PR aligns who
the projection delete targets; this one adds the missing event tombstone).
Happy to rebase over whichever lands first.

Testing

New crates/buzz-test-client/tests/e2e_workflow_def_deletion.rs (2 tests:
UUID deletion, name deletion). Verified against a local relay
(postgres/redis/minio via docker compose):

• against unfixed main: both tests FAIL (deleted def still served)
• with this change: both PASS
• e2e_long_form suite (the relay: NIP-09 a-tag deletion is a no-op for kind:30023 (and all non-workflow addressables) #714 generic NIP-33 arm): 8/8 PASS
• cargo clippy --all-targets --all-features -- -D warnings: clean
• cargo fmt --all -- --check: clean

just ci not run end-to-end locally (desktop/mobile toolchain versions on
this machine); no desktop or mobile files touched — relying on GitHub CI for
those legs.

Checklist

• cargo fmt + clippy -D warnings clean
• Regression test included (fails without the fix)
• No new unwrap() in production code paths
• No new unsafe blocks