amd bug report

[karenl03]

Description
AMD support for log collector

• amd-bug-report.sh script mimics nvidia-bug-report for AMD GPUs -- collects ROCm diagnostics, amd-smi metrics, firmware info, etc into compressed .log.gz
• K8s: AMD Daemonset with correct node affinity, toelrations, and driver namespace to deploy only on AMD GPUs
• VM: docker and native (--no-docker) installation via crusoe_watch_agent_amd.sh script

Unified log collector

• Moved log collector from separate k8s/log-collector/ and vm/log-collector/ to common/log-collector/
• log_collector.py automatically detects K8s vs VM environment and supports both NVIDIA and AMD via GPU_TYPE env var
• Two docker images: log-collector and amd-log-collector, both built by log-collector-ci (triggered on common/log-collector/*)

Jira
https://crusoe.atlassian.net/browse/CRUSOE-55851

Testing
Tested log collector capabilities on AMD, as well as NVIDIA due to new unified approach
Sample AMD bug report here

• K8s AMD: amd cluster with MI300x
• VM AMD: MI300X with Docker and Native installation
• K8S NVIDIA: ib-commandcenter-cluster and b200 cluster, generate via UI
• VM NVIDIA: run test script on L40S and A100 across all 5 images and both native/docker installation