Build a distributed lock manager simulation

• acquire(resource, clientId, { ttlMs, waitMs? }) grants a lock lease if the resource is free, returning a lease with an owner, expiry and a monotonically increasing fencing token per resource
• release(resource, clientId, token) frees the lock only if the caller still owns it with that token; releasing someone else's lock or an expired lease is rejected
• Leases expire after their TTL (injectable clock), after which another client can acquire it; the old owner's later operations are recognised as stale
• renew(resource, clientId, token, ttlMs) extends a live lease, and fails if the lease expired or was taken over
• Reentrancy: the same client acquiring a lock it already holds increments a hold count rather than blocking, and it must release as many times as it acquired
• Waiters are queued in FIFO order with an optional wait timeout; on release or expiry the first waiter is granted the lock, and a timed-out waiter is removed from the queue

Only after every requirement works.

• A protected resource that rejects writes carrying a token lower than the highest it has seen
• Deadlock detection across clients that wait on each other's locks
• Read and write lock modes