record's --target Undocumented in local-strict-review, and check's --target Help Text Stale After #1236

[ptr727]

Two small drifts surfaced by an adversarial review of #1222 (canonical_review.py record now
stamping hubCommit as a merge-base against --target), both out of scope for that PR since
fixing them needs files outside its assignment.

1. local-strict-review's "Recording the Pass" example omits --target on record

.agents/skills/local-strict-review/SKILL.md's "Recording the Pass" section gives:

python3 "$engine" status --target <target>        # JSON, take contentDigest
python3 "$engine" record --reviewer agent-skill --target <target> --expect-digest <digest> [--findings N]

for local_review.py, and its own text says a branch based on main needs --target passed
explicitly or the digest is measured against a fork point nobody read.

canonical_review.py record gained the same --target flag in #1222 (default develop), for
the same reason: hubCommit is now stamped as the merge-base against the target, and leaving
--target off silently anchors it at develop even on a branch based on main. But "The
Carried-Content Pass" section's own example still reads:

python3 scripts/canonical_review.py check --target <target>
python3 scripts/canonical_review.py record --reviewer agent-skill --unit '<key>=<digest>' [--findings N]

--target on check, not on record. Followed literally on a main-based branch, check
measures the real fork point while record stamps a merge-base against develop, which is
neither the base check measured nor a commit the reviewed content ever existed at.

Fix: add --target <target> to the record line in that example, matching the local_review.py
one above it. .agents/skills/ is hub-canonical content outside #1222's assigned file list.

2. check's own --target help string is stale after #1236

scripts/canonical_review.py's check subcommand still reads:

p_check.add_argument(
    "--target",
    default=None,
    help=f"target branch (default {DEFAULT_TARGET}), resolved as origin/<value> first",
)

local_review.target_ref (which canonical_review.py calls into) changed its resolution order
in #1236 ("Prefer an Explicitly Named Remote-Tracking Ref Over origin/ in local_review.py"): it
now tries refs/remotes/<value> before origin/<value>, so a value that is already a
remote-tracking ref (e.g. upstream/main) is used as written rather than being prefixed with
origin/. The help string's claim that it is "resolved as origin/ first" is therefore no
longer accurate. #1222 gave record's own new --target a different, accurate phrasing rather
than repeating the stale claim, which is what surfaced this.

Fix: reword check's --target help string (and cross-check local_review.py's own --target
help strings for the same drift) to match target_ref's current resolution order, documented in
its own docstring.