make help: list typecheck in the check description

[derek-miller]

Fixes the stale make help description flagged in FL-10: check gained typecheck as a prerequisite when the gate landed, but the help line still described it as format-check and lint only. One-line @echo edit, no behaviour change.

Ref: https://youtrack.dmiller.me/issue/FL-10

[svc-finitelabs]

Verified against the target itself, not just the diff.

check: format-check lint typecheck (Makefile:188) matches the new help string exactly, including the order. The edit is one @echo inside the help recipe, so there is no behaviour change and nothing downstream reads this text.

Fleet coverage is complete. FL-10 called out five repos and all five have an fl-10-help-typecheck branch open: lua-bitn#18, lua-bthome-ble#10, lua-noiseprotocol#20, lua-protobuf#17, and this one. I spot-checked the odd one the ticket flagged: lua-protobuf declares check: format-check lint check-types typecheck (Makefile:220) and its help line now reads "format-check, lint, check-types, and typecheck", so the four-prerequisite case is right too rather than being copy-pasted from the other four.

One non-blocking gap in the same block, outside FL-10's scope. all (Makefile:9, all: format lint test build) is the default target and the only real target the help text never lists. It is also the one place that runs format instead of format-check, so it rewrites the tree, and it skips typecheck altogether. If the point of FL-10 is that make help should not understate the gate, an undocumented default target that silently reformats your working tree is the same class of surprise. Say the word and I will file it; I did not fold it in here because it is a behaviour question, not a stale-string fix.

Approving. Two notes on the verdict: auto-merge was not armed when I posted, so this approval is not a merge trigger, and the Lua version matrix was still in progress at review time on f46adc8 (Check, Lua luajit-2.0, and Lua luajit-2.1 green). Nothing in a help string can move those jobs, but the branch stays blocked until they land.

[svc-finitelabs]

Adjacent drift in the same class, not covered by this PR. README.md:228 repeats the stale sentence verbatim:

make check               # Run format check and lint
make format              # Format code with stylua
make lint                # Run luacheck

So once this merges, make help will be accurate and the README will still understate the gate in exactly the way FL-10 describes. That block also omits format-check and typecheck as targets entirely, which make help does list.

If you want it folded in here rather than as a follow-up:

make check               # Run format-check, lint, and typecheck
make format              # Format code with stylua
make format-check        # Check code formatting
make lint                # Run luacheck
make typecheck           # Check annotations with lua-language-server

I checked the rest of this repo: README.md:228 is the only hit and there is no CONTRIBUTING.md here, so that one line closes it.

Fleet status for the same README drift, posted to FL-10 as well so the other four sessions do not each re-derive it: lua-bitn README.md:143 and lua-noiseprotocol README.md:152 carry the identical line, lua-bthome-ble has it at README.md:154 and 208, and lua-protobuf is already correct at README.md:165 ("Run format check, lint, check-types, and typecheck"). Protobuf is the one to leave alone, which is the reverse of the Makefile situation where it was the odd one out.

Approval above stands either way. This is a docs follow-up, not a blocker.