Fix: always restore error handler in object cache Redis _connect() (finally + \Throwable)

[vbakkenes]

Problem

Object_Cache::_connect() registers set_error_handler( 'litespeed_exception_handler' ) around the Redis connection, but only catches \Exception / \ErrorException and calls restore_error_handler() after the try/catch. If a \Error (e.g. a TypeError/ValueError on PHP 8.x) escapes the connect path, both catch blocks are skipped, restore_error_handler() never runs, and litespeed_exception_handler stays registered for the rest of the request.

Since that handler rethrows every PHP error as an ErrorException, any later unrelated PHP 8.1+ deprecation — e.g. Gravity Forms dirname(null) or Wordfence strrpos(null) — becomes a fatal Uncaught ErrorException, white-screening the whole site even with WP_DEBUG off. The stack trace points at the other plugin, so it's very hard to trace back to the object cache.

Also, the catch ( \ErrorException $e ) after catch ( \Exception $e ) was unreachable — \ErrorException extends \Exception.

Fix

Catch \Throwable (covers \Error too, and folds in the previously-unreachable \ErrorException branch) and move restore_error_handler() into a finally so the handler is always restored.

Notes

• \Throwable / finally are PHP 7.0+, within the plugin's Requires PHP: 7.2.
• Behaviour on a genuine connection failure is unchanged: $failed = true and fall back as before.

Fixes #1013