fix(masters): normalize whitespace around a non-empty Metrika counter id

[axisrow]

Summary

_metrika_counter_identity (direct_cli/browser/masters.py, added in PR #801) extracts the numeric counter id shared by the suggestion-text and read-back tag-display formats so _verify_saved can compare them for equality. The non-empty branch returned the identity unstripped:

identity = first_line.rsplit(" • ", 1)[1]
return identity if identity.strip() else text

If a real (non-empty) id has surrounding whitespace on one side but not the other — e.g. read-back "domain • 12345 \n30 целей" (trailing space before the newline) vs. suggestion "label • domain/path • 12345" (no trailing space) — the two identities ("12345 " vs "12345") wouldn't compare equal, and _verify_saved's Counter comparison would report a false save-mismatch even though the add actually succeeded.

Fix

identity = first_line.rsplit(" • ", 1)[1].strip()
return identity if identity else text

The existing empty/whitespace-only fallback behavior (PR #808's guard: return the original text unchanged when there's no real id after the last " • " separator) is unchanged — only the non-empty branch now normalizes whitespace.

Testing

• Added test_strips_trailing_whitespace_around_a_real_id, test_strips_leading_whitespace_around_a_real_id, and test_suggestion_and_whitespace_padded_tag_display_share_identity (the exact scenario from the issue: suggestion text with no padding vs. read-back tag-display text with incidental whitespace, both must normalize to the same identity) to TestMetrikaCounterIdentity in tests/test_masters.py.
• pytest tests/test_masters.py -k MetrikaCounterIdentity — 10 passed
• pytest tests/test_masters.py — 843 passed
• Full offline suite pytest — 3492 passed, 23 skipped, no regressions
• black/flake8 clean on all changed files

Closes #809

🤖 Generated with Claude Code

[axisrow]

🔍 Local review (cycle 1) — round 564e95fe-cb4c-4e02-8413-a38e69742e36

Reviewed locally (/review + Codex companion), no bots pinged.

Verdict	Reviewer	Finding	Location
SKIP (clean)	claude	Claude /review found no issues across 7 independent angles: correctness, removed-behavior, cross-caller impact, reuse, simplification/efficiency, altitude, and CLAUDE.md conventions.	—
—	codex	Codex companion review could not be programmatically parsed by its own JSON extractor on two automated attempts (prose preceded the JSON verdict block); the raw transcript was manually inspected and accepted by explicit user direction as an exception to the normal automated-parse path.	—
FIX	codex	Codex flagged (low confidence, 0.35) that stripping whitespace around a non-empty identity token also collapses distinct non-numeric malformed tokens that differ only by surrounding whitespace, partially narrowing the docstring-documented "malformed inputs must not silently match" invariant beyond the intended numeric-id case.	direct_cli/browser/masters.py:5339-5343
Applied	—	Restricted the whitespace-stripping normalization to tokens that are genuinely numeric (.isdigit()); non-numeric tokens now fall back to the original unstripped text, fully restoring the invariant for malformed input while keeping the fix for real numeric counter ids.	direct_cli/browser/masters.py

Cycle 1/3.

[axisrow]

🔍 Local review (cycle 2) — round 4243714e-0aac-4d6c-ac75-1e225fe32aef

Reviewed locally (/review + Codex companion), no bots pinged.

Verdict	Reviewer	Finding	Location
SKIP	claude	Claude /review flagged (non-blocking, altitude/design observation, not a bug) that the function has accreted three sequential defensive branches across three issues, suggesting a future single-regex rewrite might be more maintainable.	direct_cli/browser/masters.py:5307
— (clean)	codex	Manually reviewed Codex companion output (again unparseable by its own JSON extractor due to leading prose) approved the fix with no findings, independently re-verifying the #809 fix, the malformed-token guard, and a non-ASCII isdigit() edge case as non-material.	—

No FIX verdicts this round — reason to defer the design observation: the three guards each protect a distinct, tested, narrow scenario (empty token, whitespace padding, non-numeric malformed token); a regex rewrite risks silently changing behavior for one of the 12 existing regression tests without a corresponding maintainability win, and the current code passes cycle-review round 2 clean on correctness. Left as-is.

Cycle 2/3 — clean round, no FIX.

[axisrow]

📋 Review summary — all cycles

Cycle	Reviewer	Finding	Verdict	Resolution
1	(issue #809, pre-PR)	Codex found an unstripped whitespace edge case where a real numeric counter id with padding on one side but not the other would fail Counter-based save verification despite the add succeeding.	FIX	Fixed in e773a1e (original PR commit)
1	codex	Codex found stripping whitespace broadly (not restricted to numeric tokens) would collapse distinct non-numeric malformed inputs that differ only by surrounding whitespace, narrowing an existing safety invariant.	FIX	Fixed in 1ba7c44 (.isdigit() guard)
2	claude	Claude noted the function has accumulated three sequential defensive branches across three issues and floated a future single-regex rewrite as more maintainable; not a bug.	SKIP	Left as-is — see round-2 PR comment for rationale

Totals: 2 FIX (both resolved), 1 SKIP, 0 UNVERIFIED. Codex companion output was manually inspected both rounds due to a JSON-parse gap in the companion tool itself (prose preceding the JSON verdict block); accepted by explicit user direction as noted in this PR's round-1 comment.