feat(explorer): show compact counts and generate editable count queries

[2160039878-cyber]

Description

Explorer row-count badges now use compact K/M/B/T notation. SQL table menus offer Select Table Count, which opens SELECT COUNT(*) in a new editable tab. Users can add filters before running it; choosing the action never executes the count or starts table profiling.

Type of Change

• New feature (non-breaking)
• Documentation and test updates

Changes Made

• Format reported counts with the platform's compact number formatter, including promotion at rounding boundaries. Missing counts remain absent; zero remains visible. The badge still reports the provider's count, which can be an estimate.
• Reuse dialect-aware qualified-name quoting and statement terminators. The action is offered only for addressable SQL tables when a handler is available, covering both context and overflow menus.
• Wire the editable tab flow through the standalone desktop/mobile and embedded workspace shells. Existing tabs are preserved, and mobile returns to the editor. No new execution path or provider capability is introduced.

Testing

• TDD: seven component assertions and three hook cases failed before implementation.
• bun run test:components --pass-with-no-tests -t 'Studio|Sidebar|SchemaExplorer|TableItem': 382 matching tests passed, 0 failed.
• bun run test:hooks --pass-with-no-tests -t 'useTabManager': 34 passed, 0 failed.
• bun run test:unit --pass-with-no-tests -t 'generateCountQuery|generateSelectQuery|generateTableQuery|quoteIdentifier|quoteQualifiedName': 124 passed, 0 failed.
• Regressions cover large and absent counts, rounding, both menu paths, unsupported providers, missing callbacks, shell-to-row wiring, tab preservation/editing, no automatic execution, escaped identifiers and dialect terminators.
• Passed locally: format, lint, typecheck, knip, readme:check, chart:check, channels:showcase:check, security:check, production build, build:lib and attw.
• Full local tests / coverage and E2E were not completed: the Windows host lacks Helm/chart dependencies, Docker is unavailable, and existing SQLite cleanup tests encounter Windows file-lock errors. Official Linux CI verifies the full suite and 100% line-coverage gate.

Environment: Windows, Node.js 24.18.1, Bun 1.4.2. Both builds ran from a clean checkout of the submitted commit with real local dependencies.

Checklist

• Reviewed the diff and reused existing query generation and tab flows.
• Added regression tests and updated README.
• Required CI test job passes the 100% line-coverage gate on commit 10ced29.

Additional Notes

AI-assisted implementation and test execution using Codex. No new dependencies, provider changes or automatic database requests. Executing an exact count can be expensive on a large table, so execution stays under the user's existing Run control.

Current commit CI: official Linux run; 20 checks passed and the two expected Image Scan / fork SonarCloud jobs were skipped.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cevheri]

Closing as an automated bulk submission that this project cannot absorb.

This account opened 35 pull requests here in under 25 hours, a median of 17 minutes apart and four inside the same 13 seconds. That is machine cadence, not a person working through a backlog, and 35 of its 41 pull requests across all of GitHub targeted this one repository. On the earlier ones I did review, the review threads were never answered and review was never re-requested, so nothing closed out.

This pull request never reached review, and the account is now blocked at the organization level, so it cannot move forward.

If a person wants to contribute here: open one issue, wait for assignment, and send one pull request at a time.