sqlite3.Connection.iterdump() produces an unrestorable dump for a populated virtual table

Bug description

sqlite3.Connection.iterdump() emits INSERT statements for a virtual table's
own rows in addition to its schema. Those rows live in the shadow tables (which
are dumped separately), and on restore they run before the virtual table module
is instantiated, so restoring the dump fails:

import sqlite3
src = sqlite3.connect(":memory:")
src.execute("CREATE VIRTUAL TABLE t USING fts4(x)")
src.execute("INSERT INTO t(x) VALUES ('hello world')")
script = "".join(src.iterdump())
dst = sqlite3.connect(":memory:")
dst.executescript(script)
# sqlite3.OperationalError: no such table: t

The SQLite shell's .dump does not emit content rows for virtual tables.

CPython versions tested on

3.14, 3.15, 3.16

Operating systems tested on

macOS

Linked PRs

• gh-153730

首先透過 sqlite3.Connection.iterdump() 重現該範例，並將其輸出與 SQLite 的 .dump 行為進行比較。完成的標準是：來自已填入資料的虛擬表的 dump 可以在新連線上執行，而不會出現回報的 no such table 錯誤；開始前請檢查已連結的 PR gh-153730，因為這項工作已經與該 issue 關聯。