Replace CSV export with NDJSON export as the sole Shortcuts export

[sudden-break]

Summary

Removes ShortcutHealthExport (CSV format, noop_sync.txt) and makes ShortcutNdjsonExport the single Shortcuts export path for sideloaded iOS installs that can't use HealthKit directly.

Motivation

The existing CSV export produces a single flat file (noop_sync.txt) with HR,HRV,Steps,yyyy-MM-dd HH:mm lines. The NDJSON export produces 7 structured .txt files (one per data category) that are easier to parse in Shortcuts and carry richer data (sleep stages, workouts, daily metrics, calorie dedup). Having one export path simplifies maintenance and the settings UI.

What changed

New files:

• Strand/Data/ShortcutNdjsonExport.swift — core export logic (517 lines)
• StrandTests/ShortcutNdjsonExportTests.swift — unit tests (660 lines)

Deleted files:

• Strand/Data/ShortcutHealthExport.swift — CSV export (213 lines)
• StrandTests/ShortcutHealthExportTests.swift — CSV export tests (284 lines)

Modified files:

• StrandiOS/App/ShortcutExportSettingsView.swift — removed CSV card, NDJSON-only settings UI
• StrandiOS/App/StrandiOSApp.swift — removed CSV background trigger
• StrandiOS/App/RootTabView.swift — updated comment
• Strand/Data/ShortcutHealthImport.swift — updated doc comments
• Strand/Resources/Localizable.xcstrings — 64 new translation entries (8 languages)
• project.yml — updated comment

Exported files (7 NDJSON .txt files)

File	Content
heart_rate.txt	15-min windowed HR averages
hrv.txt	SDNN from R-R intervals (via HRVAnalyzer)
sleep_sleeping.txt	Sleep stages (light/deep/rem)
sleep_awake.txt	Sleep stages (wake)
workouts.txt	Workouts with Apple Health type mapping
daily_metrics.txt	RHR, HRV, resp rate, calories (deduplicated)
summary.txt	Export metadata

Behavior

• Triggered on app background (same as old CSV export)
• Opt-in via "Export for Shortcuts (Apple Health)" toggle in Settings
• Configurable 7/30/90-day lookback window
• User-chosen output folder (or Documents fallback)
• Delta export with watermark-based tracking
• Auto-migration: users with CSV export enabled get NDJSON enabled automatically
• Daily metrics: only exports completed days (excludes today)
• Calorie dedup: subtracts workout kcal from daily active_kcal

Testing

• 43 unit tests pass (all renderers + delta export behavior)
• iOS Release build compiles clean
• Unsigned IPA produced and verified installable via AltStore/SideStore
• Translations verified in compiled bundle (de, es, fr, it, pt-PT, ru, zh-Hans, zh-Hant)

What's NOT changed

• ShortcutHealthImport (import from Apple Health) — completely separate, untouched
• ShortcutExportReads protocol — moved into ShortcutNdjsonExport.swift, still provides the test seam
• WhoopStore conformance — preserved
• Background trigger — still fires on app exit, now only calls NDJSON export

[ryanbr]

Review

The NDJSON export itself is well-engineered — good test coverage (43 tests), correct delta export with watermark tracking, proper security (bookmark-based sandboxing), and the calorie dedup logic is sound.

However, removing the CSV export entirely is a breaking change for any user who has an existing Shortcut built around parsing noop_sync.txt. The auto-migration enables NDJSON, but it can't migrate the user's Shortcut workflow — their automation would silently stop working.

Suggestion: keep CSV as a legacy option and offer NDJSON alongside it.

• Make NDJSON the default for new users
• Keep the CSV toggle available (even if de-emphasized in the settings UI) so existing Shortcuts don't break
• Optionally add a deprecation note on the CSV option ("NDJSON is recommended — richer data, easier to parse")
• Remove CSV in a future version after users have had time to migrate

This way the richer format is front and center without silently breaking anyone's automation.

Minor code notes (not blockers)

1. JSON string interpolation — sport/stage values use "\(w.sport)" instead of JSONEncoder. Source values are controlled, so low risk, but JSONEncoder would be safer for defense-in-depth.
2. Midnight-boundary calorie dedup — workout calories attributed to start-day is correct by design, but a test covering a midnight-spanning workout would strengthen confidence.