fix(agentos): wait for a spawned process to start before reading its output

[eersnington]

• The client's AgentOs::spawn_process returns the pid before the sidecar has the process. The Execute request goes out from a background task.
• The read_process_output call sent ReadProcessOutputRequest right away, so a read soon after a spawn, or after a launch that failed, got ESRCH. The launch error (for example a missing command) was recorded only in the local exit state, so callers never saw it.
• Now read_process_output waits until the process has a kernel pid or a settled outcome, then reads, or returns the launch error. wait_process already waits the same way.

before: kernel error [ESRCH]: process proc-1000001-… does not exist or its replay expired
after:  kernel error [ENOENT]: command not found on native sidecar path: nosuchcmd

• The signal_process_awaited and resize_process_pty_awaited calls also send the process id before the launch lands. They are not changed here.

• Adds output_read_right_after_a_failed_spawn_returns_the_launch_error to crates/client/tests/process_e2e.rs: a read right after spawning a missing command returns ENOENT. On main it returns ESRCH.

[railway-app]

This PR was not deployed automatically as @eersnington does not have access to the Railway project.

In order to get automatic PR deploys, please add @eersnington to your workspace on Railway.

[the-company-company]

🟠 1 medium-severity finding

_{Reviewed commit b11c4ce.}

[the-company-company]

🟠 Medium · Track spawn acknowledgement separately from the optional kernel PID

kernel_pid is not a reliable acknowledgement flag: ProcessStartedResponse.pid is explicitly optional, and run_spawn accepts a successful response while only updating this watch channel for Some(pid). With a valid process_started response containing no PID, a long-running process leaves both started as None and outcome as Pending, so this call waits indefinitely even though replay is already available. Add a dedicated spawn-result/readiness watch (as the shell path does), and signal it on every successful send_execute response rather than overloading the optional PID mapping.

[eersnington]

not reachable. slop

[the-company-company]

🟠 1 medium-severity finding

_{Reviewed commit d84d3a2.}