Optimize: Update transitive closure algorithm

[hneiva]

The previous implementation scanned all edges on every iteration of a fixed-point loop, resulting in O(depth * edges) complexity. For deep dependency chains this is catastrophic - a 5000-node chain required 5000 full passes over all edges.

Replace with a standard Breadth-First Search using an adjacency list built once upfront, reducing complexity to O(vertices + edges). Benchmarks on large graphs show 22x-1333x speedups depending on graph shape, with identical results verified against the original algorithm.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.18%. Comparing base (01028b2) to head (f19b24b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #920      +/-   ##
==========================================
+ Coverage   77.14%   77.18%   +0.04%     
==========================================
  Files         130      130              
  Lines       11868    11883      +15     
  Branches     1458     1462       +4     
==========================================
+ Hits         9155     9172      +17     
+ Misses       2331     2330       -1     
+ Partials      382      381       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[bhearsum]

If you haven't already, testing this with Gecko would be a good additional sanity check (eg: that it produces identical graphs, no exceptions, etc.).

[hneiva]

If you haven't already, testing this with Gecko would be a good additional sanity check (eg: that it produces identical graphs, no exceptions, etc.).

I have tested this against gecko and found no diffs on the generated graphs.

[bhearsum]

I'm -0 on this...it adds complexity for no real world win. As long as it doesn't hurt perf in our real world cases I guess it's fine to take.

[ahal]

Seems fine to me. It's a few more lines of code, but I can't really say it's much more complex than the old algorithm