fix(cli): properly handle build log streaming during coder ping (#15600)

Closes #15584.

- The `Collecting Diagnostics` spinner now starts after the workspace
build logs (if any) have finished streaming.
- Removes network interfaces with negative MTUs from `healthsdk`
diagnostics.
- Improves the wording on diagnostics for MTUs below the 'safe' value to
indicate that direct connections may be degraded, or rendered unusable
(i.e. if every packet is dropped).

cli/ping.go

@@ -103,11 +103,6 @@ func (r *RootCmd) ping() *serpent.Command {
 			ctx, cancel := context.WithCancel(inv.Context())
 			defer cancel()
 
-			spin := spinner.New(spinner.CharSets[5], 100*time.Millisecond)
-			spin.Writer = inv.Stderr
-			spin.Suffix = pretty.Sprint(cliui.DefaultStyles.Keyword, " Collecting diagnostics...")
-			spin.Start()
-
 			notifyCtx, notifyCancel := inv.SignalNotifyContext(ctx, StopSignals...)
 			defer notifyCancel()
 
@@ -118,10 +113,15 @@ func (r *RootCmd) ping() *serpent.Command {
 				workspaceName,
 			)
 			if err != nil {
-				spin.Stop()
 				return err
 			}
 
+			// Start spinner after any build logs have finished streaming
+			spin := spinner.New(spinner.CharSets[5], 100*time.Millisecond)
+			spin.Writer = inv.Stderr
+			spin.Suffix = pretty.Sprint(cliui.DefaultStyles.Keyword, " Collecting diagnostics...")
+			spin.Start()
+
 			opts := &workspacesdk.DialAgentOptions{}
 
 			if r.verbose {