Fix O(n²) performance issue in FrameRenderer

[mcollina]

Summary

This PR fixes a significant performance issue in FrameRenderer.ts where the frame color calculation had O(n²) time complexity. The optimization reduces this to O(n) by pre-computing depth totals once instead of filtering all frames for each frame during rendering.

Changes

Performance Optimization in src/renderer/FrameRenderer.ts:

1. Pre-computation of depth totals (before the rendering loop):

   • Added totalValueByDepth Map that computes total values at each depth level in a single O(n) pass
   • This eliminates the need to filter through all frames repeatedly

2. Updated #calculateFrameColor method:

   • Changed signature to accept pre-computed totalValueByDepth map instead of the full allFrames array
   • Replaced O(n) filter operation with O(1) map lookup

Performance Impact

• Before: O(n²) complexity - for each frame, filtered through all frames to find those at the same depth
• After: O(n) complexity - single pass to build the map, then O(1) lookups during rendering
• Result: Significantly faster rendering for large flame graphs with many frames

Test Plan

• Existing functionality remains unchanged (only internal implementation optimized)
• Visual output is identical to previous behavior
• Performance improvement is most noticeable with large profile datasets

Fixes #30
Fixes #44