fix(history): an empty record is not a failed match

[ChelseaKR]

What was wrong

Found by smoke-testing the four new verbs together on main after #65 merged,
which is the check the brief asks for after a burst of merges.

data/changes/ does not exist in this repository — the watch has not recorded a
revision yet — but data/parsed/ is committed. In exactly that state:

$ ca-tariff-parse history --id smud-r-tod --all
smud-r-tod: no record matched None. That is not the same as a record that has
never changed, which is reported with one state
# History of smud-r-tod

No record matched.
$ echo $?
5

Two things wrong with that.

--all asks for nothing by name, so nothing coming back is a statement
about what has been committed, not a request that failed. Exiting 5 tells a
script the request was wrong when the record is simply empty — the same shape as
the defects this verb was written to avoid, on the other side of the fence.

{args.match!r} renders None when --all was used, so the message names a
match nobody made.

What changed

• history --all with no committed reports exits 0 and says which directory
  holds no reports, and that this is a statement about what has been committed
  rather than about whether the document has changed. --match selecting
  nothing still exits 5.
• The empty report body now says that neither an empty match nor an empty report
  set means the document has not changed — it means the committed record does
  not cover it. The old wording, "No record matched.", was the only thing a
  reader saw and it was not true for --all.

How it was verified

Full gate green: ruff check, ruff format --check, mypy (strict), pytest,
coverage 93.86% against an 85% floor. Two tests added.

The added test builds the repository's actual state — a reviewed baseline
committed with an empty changes/ beside it — rather than an empty directory,
because with neither present an earlier guard fires first and the test would
have passed without exercising this path at all.

Negative control: reverting the --all branch (if False:, read back at
cli.py:324) turns test_all_with_no_committed_reports_is_not_a_failed_match
red. Restored from a pre-edit byte copy (diff clean) and re-run green.

Prepared with AI assistance; reviewed before submission.