Update README.md

chore: clean up repo structure for public-facing release

feat: add simple EdgeRunner facade API with stream and generate methods

fix: pass dense V buffer to decode kernel for 4k-16k contexts

fix: populate dense V cache during prefill for all context lengths

fix: populate dense V cache during prefill for all context lengths

fix: dense-V decode path for turboquant quality regression

fix: disable unimplemented turboquant hybrid-V paths

merge: turboquant KV-cache integration from perf2-turboquant-isolation

chore: commit local perf2 changes before turboquant merge

perf: use dense attention for long turboquant prefill — ttft 624029 → 504410 ms with 33.24 tok/s preserved

perf: increase 16k row thinning again — 27.08 → 33.21 tok/s (+22.6%)

perf: double 16k row thinning — 24.32 → 27.08 tok/s (+11.3%)

perf: keep only top tile value row at 16k — 22.87 → 24.32 tok/s (+6.3%)

perf: fix 16k sparse score scaling — 22.33 → 22.87 tok/s (+2.4%)