WebGL Picking Pipeline Benchmark v3 — scale & concurrency

Builds on v2 by adding tests at production scale (10–1000 draw calls per frame), framebuffer size variation, and concurrent main-thread work to measure async readback's true value. Tests v2 couldn't see clearly due to single-quad noise floor.
Detecting environment...
Test scene renders here
Ready. 5 tests, ~25 measurements total.

v3 test plan

J.Draw call scaling — picking cost with 1, 10, 100, 1000 quads
K.Framebuffer size impact — 256, 512, 1024, 2048 framebuffer sizes
L.PICKING_MODE bailout at scale — bailout savings with 1, 100, 1000 quads
M.Concurrent JS work during async readback — async's real value
N.Realistic graph simulation — 1000 small quads, picking on/off
Built for LumaWeave / Sigma.js readPixels investigation. v3 focuses on scale and concurrency. bitmosh.dev