fix: check subprocess exit codes in Java tracer

[mashraf-222]

Problem

_run_java_with_graceful_timeout() in tracer.py discarded the subprocess exit code entirely. In the no-timeout path, subprocess.run() was called with check=False and the result was thrown away. In the timeout path, proc.wait() was called but proc.returncode was never checked.

If Maven or Gradle failed (compilation error, OOM, missing dependency), the tracer silently continued with missing or stale data.

Root Cause

Two bugs in _run_java_with_graceful_timeout():

• Line 28-29: subprocess.run(java_command, env=env, check=False) — no result captured
• Lines 34-47: proc.wait() called but proc.returncode never inspected

Fix

• _run_java_with_graceful_timeout() now returns int (exit code) instead of None
• Returns -1 for timeout kills (process was forcibly terminated)
• Logs warnings on non-zero exit codes
• JavaTracer.trace() handles exit codes per stage:
  • Stage 1 (JFR profiling): Warns and continues (JFR data is supplementary)
  • Stage 2 (argument capture): Raises RuntimeError (trace data is essential)

Test Coverage

10 new tests in test_tracer_exit_codes.py:

• 7 unit tests for _run_java_with_graceful_timeout() return values (success, failure, OOM, timeout, SIGTERM/SIGKILL)
• 3 integration tests for JavaTracer.trace() caller behavior (Stage 1 fail continues, Stage 2 fail raises, both succeed)

Testing

$ uv run pytest tests/test_languages/test_java/test_tracer_exit_codes.py -v
10 passed

$ uv run pytest tests/test_languages/test_java/test_java_tracer_e2e.py -v
8 passed (existing integration tests — no regressions)