docs(website): --lease waits for a starting worker (#991) (#1002)

Closes #991

### Status
Implement: done (docs only, audit 1/1) Review: round 1 passed, 0
findings Hardware: n/a Gate: running
Next: gate

- [x] The "Through a gateway" list says `--lease <id>` avoids the error:
the command waits for the worker, then runs. Checked: the added sentence
in working-with-a-device.mdx; audit and claims review traced it to
passthrough.ts:56 (lookup skipped with `--lease`), the gateway's lease
index (fleet-coordinator.ts), the worker parking requests until ready
(src/daemon/dispatch.ts:326) and the 60 s startup read cap.

## Assumptions
none

## Review

Spec review: 0 blocking, 0 fixed, 0 notes. Code review: skipped. Claims
review: 0 blocking, 0 fixed, 0 notes.
Rounds: 1. Tokens: 36.2M.
Runs: 5. Cost USD: 0.6251. Agent seconds: 218.

Assessment: complete

*Written by an agent.*

Author: V3RON

website/src/content/docs/docs/using-devices/working-with-a-device.mdx

@@ -86,7 +86,7 @@ simlock --gateway simctl --lease 3f81a2c4-9b7d-4e21-8a55-1c0e6f2d7b93.lse_9f2c6a
 - `--lease <lease-id>` names which lease the command runs under. Put it right after `simctl` or `adb`. You can leave it out when you hold exactly one lease. Against a worker it is accepted and ignored.
 - File paths in the arguments are paths on the worker, not on your machine. `simctl install` needs the app to exist on the worker already.
 - Commands that need a terminal, such as a bare `adb shell`, are refused. Input piped to the command is sent once.
-- Right after a worker restarts, the gateway hides its leases from `lease.list` and `lease.get` until the worker is no longer `starting`. That is the worker's startup checks plus one gateway refresh: usually seconds, up to about a minute, because a platform's startup read is capped at 60 seconds. In that window the command fails with `No lease is held by <your id>` although the lease stands. Do not request a new device on that message: wait a minute and run the command again.
+- Right after a worker restarts, the gateway hides its leases from `lease.list` and `lease.get` until the worker is no longer `starting`. That is the worker's startup checks plus one gateway refresh: usually seconds, up to about a minute, because a platform's startup read is capped at 60 seconds. In that window the command fails with `No lease is held by <your id>` although the lease stands. Do not request a new device on that message: wait a minute and run the command again. With `--lease <lease-id>` you do not see the error: the command skips the lookup, waits for the worker to finish starting, and then runs, up to about a minute later.
 
 ## Commands Simlock refuses