opusenc: halve the a2nlsf bisection steps

Each root was refined with 40 bisection steps, which shrinks the bracket to
about 3e-15 rad. A Q15 NLSF step is 9.6e-5 rad, so 20 steps (3e-9 rad) are
far more than the quantiser can see. The saved cos calls are the remaining
bulk of the root search after the grid table.

The coded packet streams are byte-identical to the 40-step search on 24
configurations of 90 s: speech-like and music-like input, mono and stereo,
SILK NB 8k, MB 12k, WB 16k and 32k, hybrid SWB 32k and FB 64k.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

crates/opusenc/src/silk/analysis.rs

@@ -88,8 +88,8 @@ fn cos_series(p: &[f64], w: f64) -> f64 {
 const GRID: usize = 1024;
 /// Largest multiple of ω in the series (half the LPC order, at most 16 / 2).
 const MAX_H: usize = 8;
-/// Bisection steps per root.
-const BISECT_STEPS: usize = 40;
+/// Bisection steps per root: the bracket shrinks to 3e-9 rad, far below one Q15 NLSF step (9.6e-5 rad).
+const BISECT_STEPS: usize = 20;
 
 /// `cos(m · π g / GRID)` for every grid point `g` and `m` in `0..=MAX_H`, computed with the same
 /// expression as [`cos_series`], so the table values are bit-identical to calling `cos` there.