fix: clear zombie doc-organizer indexer pid

[DevNexsler]

Summary

• treat zombie indexer.pid targets as dead instead of active
• clean stale zombie pid files during both file_status and file_index_update
• add regression coverage for zombie pid handling in status and index update paths

Root Cause

A background indexer crashed with free(): invalid next size (fast), left /data/index/indexer.pid behind, and stayed as a zombie process. os.kill(pid, 0) succeeds for zombies, so the MCP server kept reporting already_running and never allowed FTS rebuild/self-heal.

Evidence

• /data/index/indexer.pid persisted with pid 22512
• /proc/22512/status reported State: Z (zombie)
• OpenClaw doc-organizer checks returned already_running with pid 22512 repeatedly for ~24h
• indexer.log ended with free(): invalid next size (fast)

Fix

• add shared pid-resolution helper in mcp_server.py
• reap child zombies with waitpid(..., WNOHANG) when possible
• fall back to /proc/<pid>/status and treat Z as dead
• unlink stale zombie pid files before reporting/running indexer

Test Report

PYTHONPATH=. ../../.venv/bin/pytest tests/test_index_nonblocking.py tests/test_mcp_contract.py -k "index_update or file_status"
9 passed, 38 deselected, 4 warnings in 3.25s

PYTHONPATH=. ../../.venv/bin/pytest tests/test_config.py tests/test_prefect_server.py
24 passed in 1.34s