[Documentation] Explain calling of external executables in more detail

It is important to use srun or flux run when calling external executables from Python functions submitted to executorlib. Still unless nested executors are used, these jobs appear as additional tasks in the list of tasks from the queuing system. This should be explained in more detail, for example by extending the following sections:

• https://executorlib.readthedocs.io/en/latest/5-2-quantum-espresso.html
• https://executorlib.readthedocs.io/en/latest/3-hpc-job.html#nested-executors

This becomes even more important when calling GPU applications like nvidia-smi, without flux run or srun these applications have access to all GPUs of a given compute node. This can result in unintended resource sharing and slower performance. In the same way the relation to block_allocation=True should be explained in more detail.

Read the quantum-espresso page and the nested-executors section linked in the issue, focusing on how external executables are launched from submitted Python functions. Update those sections to explain srun/flux run, extra queue tasks, GPU resource access, and the relationship to block_allocation=True. Done means the documented behavior and resource implications are clear to executorlib users.