fix: iris stops working after exec a command in bash (#66)

Fix #65

when the user presses Enter to run a command, Iris sets `isCommandActive
= true` so it knows a command is running and skips rendering the
overlay. After the command finishes, `isCommandActive` needs to reset to
`false` for Iris to work again. In zsh/fish, preexec/precmd hooks send
an `IRIS_CMD_STOP` signal over the IPC pipe, which Iris receives to
reset `isCommandActive = false`. Bash has no equivalent hook, so
`IRIS_CMD_STOP` is never sent and `isCommandActive` stays stuck at true
forever

so I added a bash-only fallback inside `isExecuting()`: if
`isCommandActive` is true but `TIOCGPGRP` (a kernel API) confirms the
shell's process group is back in the foreground (meaning the command
finished), automatically reset `isCommandActive` to false. zsh/fish
still rely on IPC signals as before

also added a `PROMPT_COMMAND` hook to iris init bash to `send
IRIS_CMD_STOP` after every command (belt-and-suspenders), though that's
not the main fix

root/init.go

@@ -71,6 +71,18 @@ if [ -z "$IRIS_PID" ] && [ -z "$IRIS_RESCUE" ]; then
     export IRIS_ACTIVE_SHELL="bash"
     exec iris
 fi
+
+# Iris Autocomplete Hook
+if [ -n "$IRIS_PID" ] && [ -n "$IRIS_FD" ]; then
+  _iris_bash_precmd() {
+    printf "IRIS_CMD_STOP\x00" >&$IRIS_FD 2>/dev/null
+  }
+
+  if [[ ";$PROMPT_COMMAND;" != *";_iris_bash_precmd;"* ]]; then
+    PROMPT_COMMAND="_iris_bash_precmd${PROMPT_COMMAND:+;$PROMPT_COMMAND}"
+  fi
+fi
+
 `)
 		case "fish":
 			fmt.Printf(`