Prevent unbounded alarm scheduling promise chain in ActorSqlite.

[vaibhavSharma-cf]

When user code repeatedly moves a Durable Object alarm to a later time and scheduleRun calls to the alarm manager take longer than local SQLite commits (common in production), the alarmLaterChain promise chain grew without bound. This caused alarms to fire at times far in the past and could block SRS commits for arbitrarily long when a subsequent setAlarm moved the alarm earlier.
Replace the unbounded chain with a coalescing mechanism: at most one scheduleRun is in-flight at a time, and rapid intermediate alarm times are collapsed into a single pending value.

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[vaibhavSharma-cf]

@ask-bonk

[vaibhavSharma-cf]

Recreating as a branch on cloudflare/workerd (instead of fork) so reviewers can be added.