Update LINE Seed JP

**Font Project Git Repo URL:**
https://github.com/line/seed

**Current version:**
The published binaries carry per-weight versions: Thin 1.008, Regular 1.003,
Bold 1.010, ExtraBold 1.015 (onboarded via #10011, sourced from commit ea7a320).

**New available version:**
1.016, unified across all four weights.
Release: https://github.com/line/seed/releases/tag/v20260828
Upstream commit for METADATA.pb `source`: 0564884c67ebdb3b31e90e2106c2d061f953b6a0

**Super short description of the upgrade:**
Bug fix. The metric features `halt`, `palt`, `vhal` and `vpal` — which are GPOS-only by spec — carried hundreds of GSUB substitutions copied from `trad`, `jp78` and the vertical features. Enabling "Alternate half widths" / "Proportional metrics" in Figma or via CSS `font-feature-settings` therefore switched Japanese kanji to traditional Chinese forms (亜→亞, 医→醫, 円→圓).
The same defect was found in `vkrn`, which duplicated the `dlig` ligature rules.
All of these substitutions have been removed; glyph-form switching remains available through `trad` / `jp78` / `jp04`, which are untouched.

No glyphs, styles or instances are removed, and default rendering is unchanged.
The four weights were also versioned independently, which `opentype/family/equal_font_versions` reported as a mismatch; they are now
unified on 1.016.

fontspector `shaping/regression` tests were added under `qa/shaping_tests/` and wired into `make test` and CI, so this class of regression is now caught automatically. Against the currently published Google Fonts binaries 5 of the 6 halt/palt cases fail; against the new sources all pass.

**Bugs reported in this repo which this upgrade would fix:**
None — I searched this issue tracker for "LINE Seed" and found no existing
report of this behaviour.

Review the METADATA.pb source entry and the upstream commit 0564884c67ebdb3b31e90e2106c2d061f953b6a0, then inspect the existing qa/shaping_tests/ cases. Update the four weights to version 1.016 with the reported upstream changes, and run make test; done means the shaping regressions pass and the versions are equal.

Written by the indexing model from the issue text.