Honor external V8 root in Linux crash diagnostics #838

Problem

The exact Linux Runtime package gate can build and enter the full V8 test suite, but when a native test crashes the container wrapper does not execute its intended second ctest/GDB diagnostic pass.

Run 35510426945 (job 106077299067) exposed 35 full-V8 failures, including a common startup/teardown SIGSEGV. scripts/build-native-engine-runtime-linux-container.sh checks ICU under a hard-coded repository artifact path even when the workflow supplies --v8-root from the runner cache. The check is false, so the diagnostic block is skipped and CI retains no backtrace or failed binary.

Fix

• parse and honor the existing --v8-root argument in the Linux container wrapper;
• retain the existing build-directory variant resolution;
• on native test failure, rerun the first failing engine binary under batch GDB and print all thread backtraces;
• keep package/test exit status authoritative;
• add a shell/source contract proving a supplied external V8 root controls the ICU lookup.

Acceptance

A deliberately crashing native test on the Ubuntu 22.04 container emits a symbolized batch GDB backtrace and the job still fails. Successful builds remain unchanged and no large failure artifact is retained.