fix(page): use cached mailboxes for the initial bootstrap

[ktogias]

Summary

PageController::index() loops over every connected account synchronously in a single request to build the initial page bootstrap state, calling getMailboxes() for each one:

foreach ($mailAccounts as $mailAccount) {
    try {
        $mailboxes = $this->mailManager->getMailboxes($mailAccount); // does a live IMAP sync
        $json['mailboxes'] = $mailboxes;
    } catch (Throwable $ex) {
        $json['mailboxes'] = [];
        $json['error'] = true; // never reached on a hang, only on a thrown exception
    }
}

getMailboxes() calls mailboxSync->sync() first (a live IMAP touch) before reading the cached list. One account with a slow-to-respond or temporarily unreachable IMAP server was therefore enough to delay/block the bootstrap for every other account too, since this loop is synchronous. The existing try/catch only helps if the call throws quickly — it does nothing for a hang, since the loop never reaches the catch until the call actually returns or errors.

Fix

Adds getCachedMailboxes() to IMailManager/MailManager — a DB-only read with no live IMAP touch — and switches both account loops in PageController::index() (regular and delegated accounts) to use it. The bootstrap's speed is now fully decoupled from any account's live IMAP health. Periodic background sync jobs and opening an account's own mailbox view still do the live sync work and keep things up to date; the bootstrap just no longer needs to wait on it.

Test plan

• Updated the existing PageControllerTest::testIndex() mock expectation from getMailboxes to getCachedMailboxes.
• Added MailManagerTest::testGetCachedMailboxesDoesNotSync() confirming the new method never calls MailboxSync::sync().
• php -l clean on all changed files. I don't have a local PHP/Composer/PHPUnit environment to run the full suite in this session — happy to iterate if CI surfaces anything.
• Ran this exact change against a live production instance for several hours (an account with an intermittently unresponsive IMAP provider) — the bootstrap loads instantly regardless of that account's live IMAP state, matching the intended behavior.

[ChristophWurst]

Thanks for the patch

This also appears to be the result of agentic AI. You need to disclose that and honor CONTRIBUTING.md and AGENTS.md

Explain in your words what this change fixes. Perhaps you can demo it with before/after data flows.

[ktogias]

You're right, and I'm sorry. I missed the guideline about the disclosure.

Step by step, what this fixes:

1. On app load, PageController::index() loops through every account
   serially, calling getMailboxes() -- which does a live IMAP sync before
   returning anything.
2. So the whole page load is only as fast as the slowest account's live
   IMAP response, one at a time. A slow/large account (mine regularly takes
   20-29s and gets killed by our pool timeout) delays or hangs the entire
   app, no matter how healthy every other account is.
3. The existing try/catch doesn't help: it only catches quick exceptions,
   not a call that's just slow or hanging.

Fix: add getCachedMailboxes(), a DB-only read with no live IMAP call, and
use it for the bootstrap instead. Message listing within an open mailbox
was already cache-first and unaffected; a specific account's own folder
list still does a live sync today (separate call path, out of scope here).