docs: restructure and improve power attribution guide

This commit restructures the power attribution guide for better readability
and corrects multiple technical inaccuracies in the documentation.

Key improvements:
- Add bird's eye view section with clear analogies and core philosophy
- Provide concrete examples showing hierarchical CPU time aggregation
- Fix incorrect claim that workloads inherit power from constituents
- Clarify that each workload type calculates power independently
- Add comprehensive limitations section covering C-States, P-States, and accuracy ranges
- Include detailed code references with file paths and function names
- Add practical configuration and debugging sections

Technical corrections:
- Correct CPU time calculation hierarchy (container = sum of processes, pod = sum of containers)
- Fix multi-workload scenario to show proper process → container → pod relationships
- Replace confusing 50% CPU usage examples with clearer percentages
- Add explanation of independent attribution vs hierarchical CPU time aggregation

The guide now accurately reflects the implementation in internal/monitor/.

Signed-off-by: Sunil Thaha <[email protected]>

Author: Sunil Thaha