fix(cron): guard against NUL-byte crash in lifecycle_guard scanners

[Shunkleburger]

Summary

The terminal lifecycle_guard scanner crashes with ValueError: embedded null byte when a scanned command references a script whose decoded binary contents tokenize into a NUL-containing path. os.open() raises ValueError, but the crash sites only catch OSError — so a single bad token blocks every terminal command until the gateway recovers. This is the sibling of #76762, which fixed the resolve() path but missed the script-content path.

Root cause

_read_referenced_script() does os.open(path, flags) inside except OSError — os.open raises ValueError: embedded null byte for NUL-containing paths, which escapes uncaught.

Upstream, the crash propagates through both scanners:

• scan_for_referenced_scripts() (cron script: values) via _resolve_script_path()
• scan_command_for_gateway_lifecycle_commands() (terminal commands) via the shell-tokenized command segments — NUL bytes leak into tokens when a scanned script's binary contents are decoded

Fix (3 defensive layers, same class as #76762)

Site	Change
_iter_command_segments()	Strip \x00 at the tokenizer — covers all three scanners (path refs, bash -c payloads, launchctl detection)
_read_referenced_script()	except (OSError, ValueError) at the os.open crash site
_resolve_script_path()	NUL-strip before Path() construction (cron script: values)

NUL bytes are never valid shell characters or path components, so stripping them cannot change legitimate behavior.

Verification

• tests/hermes_cli/test_gateway_restart_loop.py — 82 passed
• Direct repro: NUL-containing path through the full scanner → no crash, safe return
• Behavior unchanged: hermes gateway restart still blocked, normal cron script values resolve identically

Test plan for CI

• Existing suite: tests/hermes_cli/test_gateway_restart_loop.py
• Recommended addition: a regression test feeding a command containing a NUL byte (e.g. bash /tmp/evil\x00path.sh) through contains_gateway_lifecycle_command_or_referenced_script() and asserting it returns False without raising.

[kshitijk4poor]

Closing as superseded by #80258, which fixes this whole bug class architecturally rather than per-callsite: path candidates are sanitized once at the ingestion boundary (NUL/empty/unexpandable tokens rejected before any OS call), text from any read_remote_script callback is sanitized at the recursion boundary (NUL = binary = nothing to scan; >1 MiB = fail closed), the remote fallback read is bounded at the source (head -c, so oversized binaries never cross the wire), and the public guard is total by construction — an unexpected walk failure logs and falls back to the direct-scan verdict instead of breaking every terminal command.

Your report and fix targeted a real member of this class — thank you. The per-callsite patches kept leaving sibling frames exposed (#76762 → #77703 → #77780 → #78256 each crashed one frame away from the previous fix), which is why we went with the boundary fix instead of merging the fragments individually. #80258 carries regression tests for the NUL-path, binary-callback, oversized-read, unset-HOME, and walk-crash cases plus an adversarial never-raises sweep.