PAI logging pipeline enhance

[Binyang2014]

There are many logging system related issue: #4865, #4694, #4615, #4468
Will try to solve these issue by this enhance:

Our goal:

1. Unified log design
2. Consider about log retrieve permission
3. Resolve log lose when node crash and disk-pressure issue

Work Item:

• Add rest API: /jobs/{job-name}/pods/{podUid}/logs P0 New get Pod logs API #5048 ETA: Code to be review before 11/4

resp:
{
   locations: [
      {name: stdout. uri: https://nodeIP/podUid/stdout.0?token=string}
      {name: stderr, uri: https://rest/v1/string……id/token?string}
   ]
}

Use this API: GET /jobs/{job-name}/pods/{podUid}/logs?tail_mode=true to retrieve the tail log

• Using multilog to replace logrotate P0 ETA: Code to be review before 11/6 Runtime log enhance openpai-runtime#31
• Redirect runtime & user log to stdout/stderr P0 ETA: Code to be review before 11/6 Runtime log enhance openpai-runtime#31
• Add permission check in log-manager P0 Add permission and new API in log-manager #5046 ETA: Code to be review before 11/6
• Pylon log-manager redirection change change pylon config for log-manager redirection #5047
• Frontend change to adapt new log API

---------------backlog---------------------

• Upload user log to Azure Blob P1
• Change res-server to auth with Azure and return the log uri P1