fix(state): treat ESRCH like ENOENT in deleted-WAL fd identity check

`_fd_is_truly_unlinked` stats `/proc/<pid>/fd/<n>` after the scan has
already read the readlink target. Between those two steps the descriptor
can be closed (ENOENT, handled by the previous commit) or the whole
process can exit (ESRCH). Both mean the descriptor can no longer keep a
retired WAL/SHM generation alive, so neither is evidence of a live holder
and neither should make the guard refuse to open the database.

Match the sibling scan in hermes_state_holders, which already skips both
errnos, by branching on `exc.errno in (ENOENT, ESRCH)`; any other OSError
still fails closed. The existing closed-descriptor test is parametrized
over both errnos, injecting the failure at `os.stat` so the ESRCH path is
exercised on every platform.

hermes_state_dbfile.py

@@ -10,6 +10,7 @@ call time, so tests that monkeypatch ``hermes_state.<name>`` keep intercepting.
 from __future__ import annotations
 
 import contextlib
+import errno
 import hashlib
 import json
 import logging
@@ -165,11 +166,12 @@ def _fd_is_truly_unlinked(fd_path: str, watched_path: str) -> bool:
     names — the guard keeps failing closed."""
     try:
         fd_stat = os.stat(fd_path)
-    except FileNotFoundError:
-        # The descriptor was closed after /proc was read. It cannot keep a
-        # retired generation alive, so do not turn this scan race into a halt.
-        return False
-    except OSError:
+    except OSError as exc:
+        # ENOENT: the descriptor was closed after /proc was read. ESRCH: the whole
+        # process exited mid-scan. Neither can keep a retired generation alive, so
+        # do not turn this scan race into a refusal (mirrors hermes_state_holders).
+        if exc.errno in (errno.ENOENT, errno.ESRCH):
+            return False
         return True
     return _identity_is_truly_unlinked((fd_stat.st_dev, fd_stat.st_ino), watched_path)