chore(mobile): upgrade to Expo SDK 57 and fix the iOS 26.5 header

[rynfar]

Adopts upstream #8609 (Expo SDK 57) and #8611 (harden native header toolbar items) from the f94a0d646..bcb855a63 review.

Stacked on #170 — #8611 needs patches/react-native-screens@4.26.2.patch, which only #8609 adds, and #8609 builds on the patch #8607 edits. Merge #170 first.

Sequencing did most of the work

Branched from origin/pylon this had 40 conflicts — 22 in the react-native-screens patch and 18 in pnpm-lock.yaml. Rebased onto the mobile branch it has 5, all in the lockfile, because #8607 had already moved the patch. Those were resolved by regenerating the lock from the merged manifests rather than hand-editing it.

The upgrade rewrites pnpm-workspace.yaml catalogs, apps/mobile/package.json, and five native patches (@expo/metro-config@57.0.12, @react-native/gradle-plugin@0.85.3, expo-modules-jsi@56.0.10, react-native-gesture-handler@2.32.0, react-native-screens@4.26.2).

Verification and its limits

vp i --frozen-lockfile passes, so the lockfile is consistent with the manifests and all five patches resolve against the versions actually locked.

This is the riskiest change set in the batch and I want to be plain about what is not verified: no simulator build, no device build, and no EAS build has been run against Expo 57. A lockfile that installs cleanly is not the same as an app that boots. This wants a mobile smoke test before it goes anywhere near a release, and it is deliberately isolated in its own PR so it can be reverted on its own.

Model: Claude Opus 5. Harness: Claude Code.

Added 2026-08-29: #8624 header overflow and back-button artifacts

Also adopts upstream #8624 (053affbed) from the f94a0d646..1f8ed54ad review, as the third commit. It corrects the header approach #8607 introduced: brand and connection status move back into headerTitle (removing BRAND_HEADER_ITEM_IDENTIFIER / renderCompactBrandHeaderItems), status width is bounded by header width minus trailing items, and the react-native-screens@4.26.2 patch gains an empty native UILabel guide so the iOS 26 scroll-edge fade works with Fabric title views. Landing #8607 without it ships the known iOS 26.5 bug where the brand rectangle morphs into the glass back button. It belongs here because it edits the 4.26.2 patch that only exists after the SDK 57 upgrade.

Clean cherry-pick. vp i --frozen-lockfile passes (only the patch content hash moved in the lock); workspace-connection-status.test.ts 8/8; mobile typecheck, native static check, lint, and format clean; review confirmed no dangling references to the removed symbols, PylonMark untouched, and every patch hunk header balanced. Still no simulator build on this branch.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.2 KiB	−109 B (−0.8%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.3 KiB	−108 B (−1.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	54.7 KiB	−822 B (−1.4%)	66.4 KiB	✅
Codex	Live turn messages	9	9	0 (0.0%)	21	✅
Claude	Total thread wire	13.3 KiB	13.2 KiB	−167 B (−1.2%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−12 B (−0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.3 KiB	−155 B (−2.3%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	55.5 KiB	−884 B (−1.5%)	66.4 KiB	✅
Claude	Live turn messages	10	9	−1 (−10.0%)	21	✅

Baseline: abe70a7 · PR result: 6465e57 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.