fix(browserstack-service): flush build-stop event on process kill (SDK-6050)

[harshit-browserstack]

Proposed changes

Fixes SDK-6050 — WDIO builds left in "Unknown Test Status" / TEST_TIMEOUT_WITH_BUILD_SUCCESS when the test process is killed mid-run (Ctrl+C / SIGINT / SIGTERM).

Root cause. The service registered only a synchronous process.on('exit', ...) handler, which Node.js restricts to sync work — it cannot await the async build-stop call. On signal, the WDIO parent exited before any stop event reached TestHub, so the backend waited out its inactivity timeout. The graceful onComplete() path always worked; the signal path was missing.

What changes. Mirrors browserstack-node-agent's BrowserStackSetup.intExitHandler + TestHubHandler.stop(signal) pattern:

1. Async signal handlers for SIGINT, SIGTERM, SIGHUP, SIGABRT, SIGQUIT (+ SIGBREAK on Windows). Each awaits the existing build-stop call with a bounded 10-second grace period, then exits with POSIX 128 + signum. Idempotent — first signal wins, subsequent signals ignored.
2. At install time, snapshot and remove pre-existing listeners for those signals so create-wdio's synchronous process.exit hook (packages/create-wdio/src/utils.ts:29) and async-exit-hook's SIGINT hook (registered transitively by @wdio/cli) cannot kill our async cleanup mid-await.
3. Thread signal through BrowserstackCLI.stop(signal) → GrpcClient.stopBinSession(signal). The StopBinSessionRequest now carries { exitSignal, exitReason: 'user_killed', exitCode: 1 }. The binary's onStop reads these and forwards them to the TestHub stop API as finished_metadata: [{ reason: 'user_killed', signal }] — matching the shape node-sdk and java-agent already send.
4. For Direct (non-CLI) mode, stopBuildUpstream(signal) adds the same finished_metadata payload directly to the PUT body.

Verification. 6-cell test matrix (Direct + CLI/BinFlow × normal / SIGINT / SIGTERM). Pre-fix kill cells: silent build, Automate status=timeout. Post-fix: build-stop PUT confirmed via binary CLI log with finished_metadata.{reason: 'user_killed', signal: 'SIGINT'|'SIGTERM'}, POSIX exit codes (130 for SIGINT, 143 for SIGTERM), status=skipped on TestHub.

Types of changes

• Bugfix (non-breaking change which fixes an issue)
• Polish
• New feature
• Breaking change
• Documentation update
• Specification changes
• Internal updates

Checklist

• I have read the CONTRIBUTING doc
• I have added tests that prove my fix is effective or that my feature works
• I have added the necessary documentation (if appropriate)
• I have added proper type definitions for new commands (if appropriate)

Backport Request

• This change is solely for v9 and doesn't need to be back-ported
• Back-ported PR at #XXXXX

Further comments

The TestHub dashboard renders build state from status_stats.in_progress, decremented by per-test TestRunFinished events. When a test process is killed mid-test, the framework (Mocha/Playwright/etc.) cannot fire its afterTest hook, so no TestRunFinished is emitted for the in-flight test. That state must be reaped on the TestHub backend when it processes a stopBuild with finished_metadata.reason='user_killed'. The SDK layer's contract ends at sending the signal; reaping in_progress test_runs is backend responsibility. This PR honors that contract and does not over-reach into per-test enumeration (matching node-sdk and java-agent behavior).

Reviewers: @webdriverio/project-committers