feat(ci): surface archive-sync failures in PR body

[ronaldtse]

Problem

The archive-sync workflow runs weekly (Wed 2am UTC) and creates automated PRs when it adds archive.org mirrors to formulas. The PR body previously said only:

Automated archive.org mirror sync.

• Chunks processed: N
• Registry: process/archive_registry.yml
• Formula updates: archive mirrors added/replaced

No detail about WHICH URLs succeeded or failed. Same failure mode as issue #225 had with formula-health.

Solution

Applies the same 3-layer pattern from PR #270 (formula-health reports):

Layer 1 — CHECKER (archive_fonts.rb)
  archive subcommand gains --json-output (appends JSONL records per URL)
  Each record: { url, formula, started_at, status, archive_url, error, completed_at }

Layer 2 — AGGREGATOR (new: archive_fonts.rb report subcommand)
  Reads all chunk-*.jsonl files
  Maps URLs back to formula paths
  Produces results/archive.json in render_report-compatible schema

Layer 3 — PUBLISHER (archive-sync.yml collect job)
  Runs render_report.rb on results/
  Embeds reports/pr-comment.md in the PR body via create-pull-request

Before / After

Before — PR body:

Automated archive.org mirror sync.
- Chunks processed: 12
- Registry: process/archive_registry.yml
- Formula updates: archive mirrors added/replaced

After — PR body includes the formatted report:

## ❯ Fontist formula checks: FAIL

| Check | Scope | Result |
|-------|-------|--------|
| archive | full · 150 items | ❌ 3 failed |

URLs — 3 failures
| Formula | Detail |
|---------|--------|
| `font_X` | https://...broken-url... · archive.org timeout |
| `font_Y` | https://...other-url... · submission returned no URL |
...

---

Automated archive.org mirror sync.
- Chunks processed: 12
...

Files

Modified:

• process/archive_fonts.rb — archive subcommand gains --json-output; new report subcommand aggregates JSONL
• .github/workflows/archive-sync.yml — each chunk writes JSONL, collect job runs report + render_report.rb, PR body includes formatted output

Test coverage

Smoke-tested locally with 3-record JSONL fixture:

• 1 newly archived URL → counted as passed
• 1 already-archived URL → counted as warning (skipped submission)
• 1 failed URL → counted as failure, rendered in failure section

Output verified: render_report.rb consumes the JSON and produces correct pr-comment.md with summary table + failure details.

Depends on

PR #270 (merged) — render_report.rb in .github/scripts/ and the unified JSON schema.