feat(import): keep the tissue loading the dive computer recorded (#1977)

[dotanalon]

Related Issue

Closes #1977

Summary

Dive computers export their own tissue state and the app dropped all of it, showing only a Buhlmann recompute that never matched the numbers the diver saw underwater. This imports what each supported format actually records and shows the computer's value wherever one exists, falling back to the app's calculation only where the computer reported nothing.

What each format carries, verified against real files:

Source	Per sample	Dive level
Garmin FIT	n2_load percent	start and end N2 load, CNS
Shearwater Cloud DB	gf99	start surface GF, end GF99, deco model, CNS
UDDF (Shearwater Cloud, Subsurface)	gradientfactor (GF99)	none
Suunto SML via Suunto Cloud	none	per-compartment N2 and He tensions at start and end (9 or 15 RGBM compartments), RGBM factors, CNS, OTU, algorithm
Ratio iX3M XML	16 compartment percentages (first and last sample stored)	start and end compartment loading, CNS

No format carries per-sample per-compartment tensions except Ratio, so the 16-compartment recompute remains the source of the tissue curves and the 3D scene.

Changes

• Profile series codec v2 adds gf99 and n2_load per-sample fields. A v1 blob still decodes with them null; an older peer stores a v2 blob and skips it until updated (existing forward-version handling), so the sync compatibility floor is unchanged.
• New dives.computer_tissue_json column (schema 224) holding a ComputerTissueSnapshot with start and end ComputerTissueState (compartment tensions in bar or loading percent, aggregate N2 load, GF99, surface GF, CNS, OTU, RGBM factors, algorithm). Tolerant JSON decode so a corrupt column cannot break dive loading.
• Garmin FIT: FitSample.n2Load, FitSummary.startN2/endN2, snapshot built in fitTissueSnapshot; wired through both the file import parser and the ImportedDive path.
• Shearwater Cloud DB: reads dive_log_records.gf99 and dive_logs header values when those optional tables exist, infers the currentTime unit per dive, aligns GF99 onto the libdivecomputer samples by nearest time (ShearwaterGf99Aligner), spells out the integer decoModel code, and prefers calculated_values_from_samples.EndGF99 over a zero placeholder in dive_details.
• UDDF: per-waypoint <gradientfactor> parsed in both readers via parseUddfGradientFactorPercent (integers as percent, fractions scaled); both UDDF writers emit it back.
• Suunto: normalizer hoists StartTissue, EndTissue and Algorithm from the cloud DiveHeader; parseSuuntoTissue accepts all three array encodings and converts Pascal to bar; the snapshot travels on DownloadedDive through the import service into the repository, where a re-import without a snapshot never clears a stored one.
• Ratio iX3M: tissueGroupNPercent read from the first and last sample into start and end states.
• Analysis: overlayComputerDecoData gains gf99Source (computer-first) and replaces the GF99 curve with the computer's samples when present; ProfileAnalysis.n2LoadCurve exposed; MetricSourceInfo.gf99Actual reports what was used.
• UI: ComputerTissueSection on the tissue card, shown only when a snapshot exists: algorithm, compartment bars or tensions in the diver's pressure unit, start versus end table, calculated value as a secondary line where both exist. New diveLog_computerTissue_* strings, l10n regenerated.
• Docs: import gap tracker row added. All 11 locales carry the new strings; non-English ones are machine-style translations that a native speaker may want to review.

Not in this PR: storing Ratio's per-sample compartment series (needs 16 new codec channels), a chart for the N2 load curve, a GF99 source badge in the legend, and a standalone Suunto SML file parser (only the cloud path is live).

Test Plan

• flutter test passes (full suite run locally; the only remaining failures are the known local-timezone tests, which pass under TZ=UTC, and the Linux ffmpeg smoke test's output-size assertion, both unrelated to this change)
• flutter analyze passes
• Manual testing on: not manually tested

Fixture-backed tests cover the real Garmin Descent trimix file (start 0, end 86, peak 191 N2 load), the Shearwater Cloud CCR export (end GF99 62), the Perdix 2 and Petrel 3 UDDF exports (GF99 on 195 of 385 samples), and the Ratio iX3M export (16 compartments). The Shearwater fixture's sample table is empty, so per-sample GF99 alignment is covered by synthetic SQLite databases only. Suunto tissue parsing is covered by unit tests built from real EON Core and HelO2 header values; no fixture in the repo carries the tissue block.

Screenshots

Not captured; the new tissue card section renders only for dives that carry an imported snapshot, and no such dive exists in the test fixtures that run in a widget harness.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

📦 Build artifacts for this PR · commit f0c659e

Platform	Download
Android (APK)	android-apk
macOS	macos-build
Windows	windows-build
Linux	linux-build

Artifacts expire in 7 days. Downloading requires being signed in to GitHub. macOS needs two extractions: unzip the downloaded artifact, then unzip the submersion-macos.zip inside it to get a runnable submersion.app. The build is ad-hoc signed — right-click → Open on first launch.

_{Updated automatically on each push.}

[dotanalon]

@ericgriffin can we merge this?