fix(core): handle shutdown lock cleanup rejection

[realfishsam]

Summary

• make signal shutdown handler synchronous so Node does not ignore a returned rejecting Promise
• catch lock-file cleanup failures, log them, and still exit with the intended success code
• guard against duplicate SIGTERM/SIGINT shutdown execution

Fixes #720

Test Plan

• npm run build --workspace=pmxt-core

[realfishsam]

PR Review: PASS (NOT VERIFIED)

What This Does

Makes server shutdown lock-file cleanup rejection-safe and idempotent so signal handling still exits after cleanup failures.

Blast Radius

Core server startup/shutdown path only.

Consumer Verification

Before (base branch):
Base shutdown awaited lockFile.remove() directly before process.exit(0), so a rejected cleanup could prevent clean process exit.

After (PR branch):
PR uses a shuttingDown guard and always exits in .finally() after logging cleanup errors. Core build/Jest passed (24 suites, 644 tests); root verification then hit missing pytest.

Test Results

• Build: PASS
• Unit tests: CORE PASS (24 suites, 644 tests; root verification blocked by missing pytest)
• Server starts: PASS during root verification
• E2E smoke: NOT VERIFIED with signal-level process test

Findings

No blocking findings.

PMXT Pipeline Check

• Field propagation (3-layer): N/A
• OpenAPI sync: N/A
• Financial precision: N/A
• Type safety: OK
• Auth safety: N/A

Semver Impact

patch -- server shutdown robustness fix.

Risk

Actual SIGINT/SIGTERM behavior with a failing lock-file removal was not reproduced; conclusion is from code path and passing build/tests.