Default untyped bind parameters to string and tidy Statement::__clone()

[tyrsson]

Cherry-picks 588e5d6 onto 0.5.x.

Why

PR #78 ("Fix argument count error") was based on mago-format-reformat, not 0.5.x. That branch had already been merged into 0.5.x via #57 on 2026-08-31 — five weeks before #78 landed on it on 2026-10-07. So #78's merge commit (2f6aeb5) is not an ancestor of 0.5.x and the fix never reached the default branch, while issue #83 was closed and placed on milestone 0.5.0.

Net effect: #83 was recorded as fixed in 0.5.0 but the bug was still live on 0.5.x. A check of all 33 merged PRs confirms #78 is the only one whose commits are not on 0.5.x (#58 and #60 landed correctly via #57).

What

bindParametersFromContainer() only appended a type character for parameters carrying errata, so untyped parameters produced an empty or short types string and bind_param() failed. Untyped parameters now bind as s.

__clone() is restored to the mysqli Statement — 0.5.x currently has no __clone() at all, so a cloned statement shared its ParameterContainer. It now deep-clones the container, resets isPrepared, and gains a void return type.

Testing

• phpunit --testsuite "unit test" — 193 tests, 385 assertions, OK.
• Both new unit tests were run against origin/0.5.x's Statement.php and fail there, so they genuinely cover the fix.
• mago format --check, mago lint and mago analyze are all clean (121 baselined issues filtered, no new findings).
• test/integration/Mysqli/StatementResultTest.php needs a live MySQL, so it is left to CI.

Cherry-picked from 588e5d6e06cc1874d806fb008520cd0b034b0455; the resulting tree is byte-identical to that commit. The only conflict was the absent __clone().

Closes #83

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!