feat: Add ability to GroupBy Samples by line number and relative line number to function start

[simonswine]

Feature Request

Summary

Add the ability to group profiling samples by more granular code location attributes, specifically:

• Absolute line number within a file
• Relative line number from function start

Motivation

Currently, samples are primarily grouped by function/method names. However, for complex functions with multiple code paths, it would be valuable to analyze performance at a more granular level by grouping samples by specific line numbers within those functions.

This would enable users to:

• Identify performance hotspots within large functions
• Analyze specific code paths or branches more precisely
• Understand which lines within a function consume the most resources
• Compare performance across different sections of the same function

Prior Art: pprof Granularity Options

The pprof tool already supports multiple granularity levels for aggregation:

granularity
  -functions       Aggregate at the function level.
  -filefunctions   Aggregate at the function level.
  -files           Aggregate at the file level.
  -lines           Aggregate at the source code line level.
  -addresses       Aggregate at the address level.

Pyroscope could provide similar GroupBy options, particularly implementing the -lines equivalent for line-level aggregation.

Use Cases

1. Long function optimization: Identify which sections of a large function are the bottleneck
2. Loop analysis: See which iterations or branches within loops consume the most time
3. Conditional path analysis: Understand performance differences between if/else branches
4. Source code correlation: Make it easier to map profiling data directly to specific source lines

Additional Considerations

• UI/UX considerations for displaying line-level granularity
• Performance implications of querying at line-level granularity
• Could also consider file-level and address-level grouping similar to pprof