fix: keep keyword billing refusal reasons

[hamitarsln]

What and why

Keyword planning called every billing refusal a spending limit. A missing model price or an already attempted operation could therefore get the wrong explanation.

Keep actual run/project limits as spending limits. Other billing refusals keep their fixed code, and the failure message names it. Raw exception text stays private. Existing call replay and spending guards stay in place.

Verification

• Six non-limit regression cases failed before the fix.
• 163 keyword-plan tests passed; 11 tests skipped.
• Tests cover real limit codes, pricing/recovery refusals, safe error text and retries making only one model call.
• Ruff, format and diff checks passed.

Notes

Verification uses local test providers. The changed worker has not been deployed or tested with a live billing refusal. No pricing or spending authority changed.

[Copilot]

🟢 Approval recommended

The implementation matches the stated billing, privacy, and replay requirements with focused regression coverage.

0 open findings

What changed in this PR

Preserves specific billing refusal codes while keeping sensitive diagnostics private and preventing duplicate model calls.

Changes:

• Distinguishes spending limits from other billing refusals.
• Adds safe, cause-specific failure messages and regression coverage.

File	Description
src/​tin_lite/​keyword_plan_activities.py	Preserves and reports safe billing refusal codes.
tests/​test_keyword_plan.py	Tests refusal classification, privacy, and retry behavior.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[demegire]

Superseded by #403: keyword plan billing refusals now use the shared billing_refusal wording, so only real run/project limits read as spending limits.