feat(learning): persist roadmap progression locally without an account

[OthmaneZ05]

Summary of Changes

A roadmap is played across several sessions, but until now every player state died with the page: reloading meant starting from step 1. This PR persists roadmap progression locally — completed steps, validation attempts and revealed hints — with zero account and zero auth, so reopening a roadmap resumes exactly where the learner left off, even after fully restarting the app.

• Backend store (~/.torollo/progress.json, next to projects.json): new ProgressService — versioned envelope (version: 1), one entry per (projectId, roadmapId), per-step data keyed by stable step ids (reordering/re-editing a roadmap file can't corrupt progress; translations share progress). Write-then-rename so a crash can't truncate the file. A corrupt or unknown-version store is moved aside as progress.json.corrupt, the app starts fresh, and the UI tells the user (one-shot storeRecovered flag) — never a crash, never silent.
• Recording: POST /validate now records the attempt + latest verdict at the API layer (the engine stays stateless; a storage failure is logged and never blocks the verdict). Hint reveals are pushed by the player as absolute counts (fire-and-forget, self-healing). New endpoints: GET/PUT …/hints/DELETE under /api/learning/progress/:projectId/:roadmapId.
• Player: opening a roadmap hydrates progress and resumes on the first incomplete step; restored steps show a ✓ marker in the step list (validator results are deliberately not replayed — they describe a past container state). New "Restart roadmap" action behind the same light two-click brake as the solution reveal. Recovery notice (amber, dismissible) when a corrupt store had to be reset. i18n en/fr.
• Lifecycle: deleting a project purges its progress entries.

Contract documented in docs/learning-api.md (store format = the future migration contract).

Types of Changes

• New feature / node type addition
• Bug fix (non-breaking change resolving an issue)
• Refactoring / structural cleanup
• Documentation update

Verification & Testing

Automated Checks

• Run npm run lint successfully with no errors
• Run npm run build successfully with no compilation errors
• Run npm test successfully (all tests pass)

Backend: 237 tests (20 new — ProgressService round-trips on real tmp files incl. corruption/unknown-version recovery and reset scoping; controller contract incl. "recording failure never blocks the verdict"). Frontend: 225 tests (12 new — hydration, first-incomplete-step resume, hint push, reset, recovery notice). Note: a few heavy pre-existing modal/canvas tests sit close to their 5s timeout and can flake under parallel load on a busy machine (seen on main files untouched by this PR, e.g. PostgresModal/NoSqlModal/CanvasPage); they pass consistently when the machine isn't saturated.

Manual Verification

Full kill & restart scenario against real Docker (Playwright + Chrome, example roadmap):

1. Completed steps 1–2 through the UI (real web + db containers, seeded users table), revealed step 3's hint.
2. Killed both dev servers, restarted → roadmap reopens on step 3, steps 1–2 carry ✓ markers, step 3's hint is still revealed; progress.json holds the exact attempts/verdicts.
3. "Restart roadmap": first click arms the confirmation, second resets to step 1 and deletes only that roadmap's entry. FR labels verified.
4. Hand-corrupted progress.json → app opens normally, amber notice shown once and dismissible, original file kept as progress.json.corrupt, next validation recreates a fresh store.
5. Deleted the project → its progress entries purged with it.

Checklist

• My code follows the repository's code style and lint standards
• I have updated the documentation or instructions if necessary
• All unit and integration tests are passing