[zmuxuny]

Which Issue(s) This PR Fixes

Fixes #7098
Fixes #7126
Fixes #7149
Fixes #7162

Brief Description

Consolidates related frontend accessibility fixes into one review, superseding #7104, #7130 and #7150 while retaining the shared instance-selector fix.

• Localized queue-specific names for topic, offset and close-detail controls.
• Stable Grafana-dashboard and alert-rule catalog filter names during search.
• Named instance inventory search/type filters after keyboard selection and clearing.
• Shared instance selector names during search, selection, clearing and loading failure.

Tests exercise real Ant Design controls in Chinese and English. No backend behavior or dependency change.

Validation

Fresh combined local run: 10 files, 122 tests passed. Focused regressions: 14 passed across 5 files; the same 14 fail on unchanged base production (79 unrelated tests skipped). TypeScript build, ESLint, Prettier and git diff --check pass; ESLint has one pre-existing QueueBrowser fast-refresh warning.

Combined command from web/:
npx vitest run src/components/__tests__/QueueBrowser.test.tsx src/components/__tests__/AlertRuleAssetList.test.tsx src/components/__tests__/GrafanaDashboardList.test.tsx src/components/__tests__/InstanceSelect.test.tsx src/pages/instance/__tests__/InstancePage.test.tsx src/services/alertRuleAssetService.test.ts src/services/grafanaService.test.ts src/i18n --pool=threads --maxWorkers=1

The jsdom run emits existing pseudo-element getComputedStyle and navigation-not-implemented warnings and exits successfully. No live cluster or manual screen-reader session; no claim of full frontend or hosted CI passing.

AI assistance was used for investigation, implementation, consolidation and verification.

[zmuxuny]

CI follow-up: this PR and #7102, #7113, #7215, #7143, and #7114 now each carry the same five-line ConsumerPage test-isolation fix from 381c5c4 (previously validated in #6758, with credit to pilichoumao's #6521). The failing manual stack/subscription tests could race the real two-second background polling and consume mock responses. Only those two tests now fake setInterval/clearInterval, and real timers are restored after cleanup. Production polling, all existing assertions, and the independent real two-second auto-refresh test are unchanged. No business changes were combined across PRs.

Validation: all six branches passed their focused suites, changed-file ESLint/Prettier, a forced TypeScript build, production build, and license checks. Focused results were 181 tests here, 146 in #7102, 381 in #7113, 85 in #7215, and 59 each in #7143/#7114. This PR additionally passed the complete frontend suite: 162 files / 1,845 tests. Backend Java/Go code was untouched and was not rerun locally. New hosted CI is queued, not yet green: https://github.com/apache/rocketmq-dashboard/actions/runs/39941472643 .

AI assistance was used for diagnosis, patch verification, and validation.

[zmuxuny]

@lizhimins Following up with the completed CI results: #7163, #7102, #7113, #7215, #7143 and #7114 now all have successful CI runs for their current heads. This PR's run is https://github.com/apache/rocketmq-dashboard/actions/runs/39941472643 . The shared ConsumerPage test-isolation fix described above is included; the six business changes remain separate.

Could you review this batch and re-review #5251 and #5025, whose current-head CI also passes? Those two still show your earlier changes-requested reviews. Please flag any remaining review items, and consider merging the PRs that meet the project's requirements. If another maintainer should review part of the queue, a reviewer or preferred order would help. Keeping the request in one thread rather than pinging every PR. Thanks.