Replica identity full with no primary key

Duplicate rows are not replicated to clickhouse. Updates are not propogated.

```
┌────────────────────────────────────────────────┬────────────────┬───────────────────────────────┬────────────────────┐
│ Operation (Postgres) │ Postgres state │ ClickHouse FINAL live │ Correct? │
├────────────────────────────────────────────────┼────────────────┼───────────────────────────────┼────────────────────┤
│ INSERT (1,'x'), (1,'x') — separate txns │ 2 rows │ 2 rows │ ✅ │
├────────────────────────────────────────────────┼────────────────┼───────────────────────────────┼────────────────────┤
│ INSERT (1,'x'), (1,'x') — same txn / bootstrap │ 2 rows │ 1 row │ ❌ collapse │
├────────────────────────────────────────────────┼────────────────┼───────────────────────────────┼────────────────────┤
│ UPDATE both → (1,'y') │ 2 rows │ 4 rows (1,x ×2 + 1,y ×2) │ ❌ duplicates │
├────────────────────────────────────────────────┼────────────────┼───────────────────────────────┼────────────────────┤
│ DELETE both │ 0 rows │ 4 rows (2 tombstones ignored) │ ❌ nothing removed │
└────────────────────────────────────────────────┴────────────────┴───────────────────────────────┴────────────────────┘

```

No files, tests, or entry points are identified in the issue. Start by reproducing the listed duplicate-row, update, and delete cases for PostgreSQL replica identity full with ClickHouse, then trace the replication path that handles these changes. Done means duplicate rows remain distinct and both updates and deletes produce the expected ClickHouse state.

Written by the indexing model from the issue text.