16368791

By: ValentinHirschi <valentin.hirschi@gmail.com>

add a smooth compact threshold localization profile

Provide the opt-in runtime uv_localisation.smooth_sliver setting in the
existing scalar and dual evaluators. Preserve hard-sliver defaults, unit
pole value and even signed radial damping; the compact bump has vanishing
boundary derivatives. Keep settings serialization and completion schemas
aligned.

Validation: nine focused tests, workspace/all-target check and clippy pass.
Tests include the actual generated helper's zero PV with its negative-radius
mirror. Full GL638 boundary/interior controls retain approximately lambda^-1
cut-summed behavior; strict Arb rotation flags remain. This profile fixes the
observed hard-boundary strip but worsens the tested MC pilot, so it stays
opt-in and does not claim a global sampling improvement.

User prompts:
Continue as planned, but remember you may want to change the localization "h" function that multiplies the local threshold CTs. Also the idea of imposing a sliver (i.e. a hard theta function) may help perhaps.
Remember though that these multiplier functions must always be symmetric around r^\star so that the PV (real part) of the integrated threshold CT remains zero (otherwise current implementation breaks).

Continue as planned, but look into the current implementation, it should already offer such flexibility.
But feel free to add some if you need be (but make sure it's steered through runtime integrand toml parameters so the user can easily access it). And it's only weakly tested, so if you find any bug there you can also adjust it.

Timeline