fix: preserve stderr output when stdout is not empty during check runs

[YasserYG8]

Description:

In becwright.engine._run_check, the engine captures and processes stdout/stderr of check subprocesses using a short-circuiting logical or:

output = proc.stdout.strip() or proc.stderr.strip()

If a check command prints standard logging or status to stdout but writes error tracebacks, linter warning details, or debug messages to stderr, proc.stderr is silently ignored and discarded. As a result, the developer sees that the commit is blocked (from the non-zero exit code), but has no error detail in the output since it was printed to stderr.

---

Steps to Reproduce

1. Define a rule that calls a custom script.
2. Make the custom script print to both standard output and standard error:

   import sys
   print("Initializing check...")
   print("Error: division by zero! Traceback...", file=sys.stderr)
   sys.exit(1)

3. Run becwright check.
4. Observe that the check blocks, but ONLY prints Initializing check.... The error message is completely lost.

---

Proposed Fix

Modify src/becwright/engine.py to concatenate both stdout and stderr outputs if they are present, rather than discarding stderr:

outputs = [proc.stdout.strip(), proc.stderr.strip()]
output = "\n".join(o for o in outputs if o).strip()