fix(supervisor): stop selecting closed shutdown channel

[theolujay]

Problem

After SIGTERM, Supervisor.Run continues selecting on the closed gracefulShutdownC channel. Since receives from a closed channel are always ready, the supervisor loop spins throughout the graceful shutdown period and consumes about one CPU core.

Change

Disable the supervisor's graceful-shutdown select case after its first notification.

Outcome

Observed during a local test with a long-lived SSE request:

State	Before	After
Idle	~0% CPU	~0% CPU
Streaming	~0% CPU	~0% CPU
Draining after SIGTERM	~100% of one core	~0% CPU

The SSE request completed after SIGTERM, and cloudflared exited after draining.

Reproduction

1. Run a local origin that streams one SSE event per second for 90 seconds:

   package main
   
   import (
       "fmt"
       "net/http"
       "time"
   )
   
   func main() {
       http.HandleFunc("/", func(w http.ResponseWriter, r *http.Request) {
           w.Header().Set("Content-Type", "text/event-stream")
           w.Header().Set("Cache-Control", "no-cache")
           w.Header().Set("Connection", "keep-alive")
           w.WriteHeader(http.StatusOK)
   
           flusher, ok := w.(http.Flusher)
           if !ok {
               return
           }
   
           for i := range 90 {
               if _, err := fmt.Fprintf(w, "data: event %d\n\n", i+1); err != nil {
                   return
               }
               flusher.Flush()
   
               select {
               case <-r.Context().Done():
                   return
               case <-time.After(time.Second):
               }
           }
       })
   
       if err := http.ListenAndServe("127.0.0.1:8000", nil); err != nil {
           panic(err)
       }
   }

2. Start cloudflared with a Quick Tunnel and its metrics/pprof listener:

   ./cloudflared tunnel \
     --url http://localhost:8000 \
     --grace-period 100s \
     --no-autoupdate \
     --metrics 127.0.0.1:20241

3. Open the generated trycloudflare.com URL in another terminal, keeping the stream open:

   curl -N https://generated-name.trycloudflare.com/

4. Send SIGTERM to the cloudflared process while the stream is active:

   kill -TERM <cloudflared-pid>

5. Observe CPU during the drain with top or by sampling the process CPU time. Before the fix, cloudflared used about one core until shutdown completed; after the fix, it stayed near idle.

6. Inspect goroutine state during the drain:

   curl 'http://127.0.0.1:20241/debug/pprof/goroutine?debug=2'

   Before the fix, the supervisor goroutine was runnable in Supervisor.Run. After the fix, it was blocked in the main select while waiting for shutdown work to finish.

---

Closes #1753

I acknowledge PR #1755 covers the same fix and I'm comfortable with it.