fix(renderer): restore quality gate baseline

[SunwardL]

Summary / 概要

Restore the renderer quality-gate baseline on the current develop branch by fixing two independent stale representations:

1. Express the existing ANSI escape-sequence matcher in a lint-safe form.
2. Align UsageQuotaPanel test expectations with the new fixed 12-week daily-history design.

恢复当前 develop 的 renderer 质量门禁基线：一项修复 lint 表达方式，另一项修复 Usage 12 周历史重构后遗漏更新的测试断言。

Why / 背景

1. Intentional ANSI matching triggered no-control-regex

stripAnsiSequences() intentionally matches ESC (U+001B) so terminal color and control sequences are removed before background-shell output is rendered.

The previous regex literal was functionally correct, but ESLint's no-control-regex rule rejects control characters written in regex literals. This caused the full npm run lint quality gate to fail even though the runtime behavior was correct.

The pattern is now constructed with new RegExp() and String.fromCharCode(27), following the existing convention in src/main/claude-subscription-auth.ts. The pattern branches and global matching behavior remain unchanged.

2. The 12-week Usage history refactor left stale range assertions

Upstream commit 586283f3 introduced SidebarUsageHistoryCard and intentionally changed daily usage loading to a fixed HISTORY_RANGE_DAYS = 84, allowing the card to render 12 weeks × 7 days.

The range selector now has two distinct responsibilities:

• Daily history always loads 84 days for the fixed 12-week card.
• Model usage continues to follow the selected 7d, 30d, 90d, or all range.

The existing UsageQuotaPanel.test.ts still expected both requests to follow the selector. The first #1264 quality-gate run therefore failed at the daily-range assertion with:

expected 84 to be 7

This was a deterministic stale-test failure in the current base rather than a runtime regression caused by the ANSI change.

Failure evidence:
https://github.com/KunAgent/Kun/actions/runs/33708795909/job/100503661816?pr=1264

Changes / 变更

• Replace the ANSI regex literal with an equivalent lint-safe RegExp construction.
• Preserve ANSI stripping behavior without adding dependencies.
• Update all four daily-history expectations to 84 days.
• Keep model-usage expectations unchanged at 7, 30, 90, and 365 days.
• Keep the change intentionally narrow: 2 files, no UI, layout, or product-behavior change.

Media / 截图或录屏

Not applicable. There is no visual or interaction change.

Tests / 测试

Focused tests:

• BackgroundShellOverlay.test.ts
• UsageQuotaPanel.test.ts
• Result: 14/14 passed

Coverage includes:

• ANSI sequence stripping.
• Plain-text preservation.
• Fixed 84-day daily history.
• Selector-driven model ranges.
• Equivalent output between the old and new ANSI patterns for representative plain text, SGR color, and erase-line sequences.

Validation / 验证

• I agree that this contribution is submitted under the Contributor License Agreement. / 我同意本贡献遵循 Contributor License Agreement 提交。
• npm run test — the local full-suite command did not return a conclusive exit code within the available execution window; GitHub Actions is the authoritative full-suite validation.
• npm run typecheck
• npm run build — not required because no runtime or packaging behavior changed.
• npm run dev — not applicable because no runtime/UI behavior changed.
• UI change: video or GIF attached — not applicable.
• Logic change: existing unit-test expectations were updated to match the intended 12-week history behavior.
• npm run lint — passed with 0 errors and 30 pre-existing warnings.
• git diff --check

Notes / 备注

This PR contains only renderer quality-gate baseline repairs. It remains separate from the Memory/P1 work so the feature PR stays focused.

After CI identified the stale Usage assertion, the original #1264 commit was amended into one clean commit:

ff1e1ed16478c1cd43aba6e1c37f9da15a5ddc39