fix(test): scrub personal PII from provider fixtures + add recurrence guard

[plauzy]

Summary

Removes personal email addresses (PII) that were committed into provider test
fixtures, and adds a guard so it cannot happen again.

• Deleted 9 orphaned test/providers/fixtures/antigravity_cli_*.txt fixtures
  that captured a maintainer's Antigravity/Google login banner
  (…@gmail.com (Google AI Ultra)). These files are referenced by no test —
  the antigravity suite loads the separate agy_*.txt fixtures — so removing
  them has zero test impact.
• Redacted a second contributor's account email in
  test/providers/fixtures/claude_code_new_tui_completed_raw.txt (which is
  used by test_claude_code_unit.py) to user@example.com. The address sat in
  the header banner; the test only asserts COMPLETED status and the bottom-box
  pattern, so status detection is unaffected.
• Added test/test_fixtures_no_personal_pii.py — a fail-closed guard that
  scans every test/**/fixtures/** file for personal-provider email addresses
  and fails CI if any are found.

No personal email remains in any tracked file after this change.

How this happened

Provider status-detection fixtures are recorded from live CLI TUIs. Several
CLIs (Antigravity, Claude Code) print the authenticated account's email in
their startup/login banner. When those raw terminal captures were saved verbatim
as fixtures and committed, the account emails rode along inside the ANSI escape
stream — not obvious in a rendered terminal, but plainly visible in the file and
in the PR diff. One set shipped in #436 (which merged before the removal landed);
another had been present in an older claude_code capture.

Because the addresses are embedded among ANSI escape codes, they slipped past
manual diff review.

Prevention

• Automated guard (this PR): test/test_fixtures_no_personal_pii.py fails CI
  the moment a fixture contains an address at a personal mail provider (gmail,
  icloud, outlook, proton, …). Synthetic/tooling addresses (@example.com,
  git@github.com) are intentionally allowed so legitimate fixtures still pass.
• Fixture-recording hygiene: when capturing a new live-TUI fixture, scrub the
  login/banner line before committing — replace the account email with
  user@example.com. Prefer capturing on a throwaway/synthetic account where
  possible.
• History note: because feat(agui): AG-UI protocol adapter + generative UI (L1, default-off) — supersedes #387 #436 already merged, the removed addresses remain in
  git history at the merge commit. If full history scrubbing is desired, that
  requires a maintainer-run git filter-repo/BFG pass on the default branch
  (out of scope for a normal PR); the affected account owners may also wish to
  treat the addresses as publicly exposed.

Testing

• test/test_fixtures_no_personal_pii.py — passes (no PII remains).
• test_claude_code_unit.py, test_antigravity_cli_unit.py,
  test_tool_mapping.py — pass (redaction/deletion broke nothing).
• black/isort clean.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@a9af159). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #456   +/-   ##
=======================================
  Coverage        ?   89.35%           
=======================================
  Files           ?      156           
  Lines           ?    18510           
  Branches        ?        0           
=======================================
  Hits            ?    16539           
  Misses          ?     1971           
  Partials        ?        0           

Flag	Coverage Δ
unittests	89.35% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[fanhongy]

LGTM

[plauzy]

@fanhongy TY for fast approval - i created #457 in light of the issue which we can all discuss at a later point.

This issue was low-severity but probably a good reason to think about how we would deal with a more severe issue if/when that happens. cc @haofeif @call-me-ram