## Implementation Plan: DeepSearch Playwright Async Context Manager Race Fix

### Changes Required

**File:** `devplacepy/services/jobs/deepsearch/crawl.py`

1. **Add import** for `asyncio` at module top (if not already present).
2. **Add module-level lock** immediately after imports:
   ```python
   _pw_lock = asyncio.Lock()
   ```
3. **Guard the Playwright call** inside `fetch_page()` (around line 240). Replace the bare invocation with an `async with` block that acquires the lock:
   ```python
   if len(text) < MIN_PAGE_CHARS:
       async with _pw_lock:
           r_title, r_text, r_status, r_links = await _render_with_playwright(url)
   ```

No other files require modification. The change is minimal (adds 2 lines of new code, modifies 1 line) and serializes only the Playwright context manager entry, preserving HTTPX concurrency.

### Verification Commands

The project’s test and lint commands must be run inside a **Python virtual environment** to avoid the PEP 668 system-package restriction. If a virtual environment does not exist, create and activate one first:

```bash
python3 -m venv .venv && source .venv/bin/activate
pip install --upgrade pip
```

Then execute:

```bash
pip install -e '.[dev]' -q && make test-unit
```

And:

```bash
pip install -q ruff && ruff check .
```

### Definition of Done

- [ ] The code change (lock addition + guarded call) is applied to `crawl.py`.
- [ ] `ruff check .` exits with code 0 (no lint errors).
- [ ] `make test-unit` exits with code 0 (all unit tests pass).
- [ ] No existing functionality is broken – the same tests that passed before still pass.
- [ ] (Optional) A manual test confirms that two concurrent `fetch_page` calls both needing Playwright no longer produce the `__aexit__` error.