NNI not running anymore without error messages when CPU reached 100% once

[BitCalSaul]

Describe the issue: Once the CPU utilization reached 100% once, even though NNI will finish the running trials but will not run the remaining trials.

Environment:

• NNI version: 3.0
• Training service (local|remote|pai|aml|etc): local
• Client OS: ubuntu 22.04, 20.04
• Server OS (for remote mode only):
• Python version: 3.8
• PyTorch/TensorFlow version: 2.1.0
• Is conda/virtualenv/venv used?: conda
• Is running in Docker?: no

How to reproduce it?: You could run a task that consumes CPU resources across multiple trials simultaneously, and you will observe this issue.

I think this issue is as the same as this one #965 .