fix(cli): show help for bare exec command

[jdx]

Fixes #1360.

exec intentionally owns -h and --help so a fully specified invocation can render the wrapped script's usage spec. Handle the bare command before required positional validation so usage exec --help prints the exec command page instead of reporting a missing <COMMAND>.

This preserves wrapped-script help for usage exec <COMMAND> <BIN> --help and supports the x alias plus both help spellings.

Tests:

• cargo test -p usage-cli --all-features
• cargo clippy -p usage-cli --all-features --all-targets -- -D warnings

This PR was generated by Codex.

---

Note

Low Risk
Small pre-parse branch in the CLI entrypoint; behavior change is limited to bare exec help and is covered by new tests.

Overview
usage exec -h / usage exec --help (and the x alias) now print the exec subcommand help instead of failing on missing <COMMAND> / <BIN>.

Cli::run intercepts that two-argument pattern before normal parsing and renders the exec page via usage_rs::help::render_styled, matching what usage help exec already shows. Invocations with a command and binary still route -h/--help to the wrapped script’s usage, unchanged.

Integration test test_exec_help_without_a_script covers exec/x × -h/--help and expects success with empty stderr.

^{Reviewed by Cursor Bugbot for commit be4fdc1. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Instruction counts

benchmark	trend	instructions	Δ	wall (min)	Δ
markdown	█▁	341,374,535 → 341,326,870	-0.01%	57.26 → 57.18ms	-0.14%
startup	▁█	903,091 → 903,841	+0.08%	1.15 → 1.16ms	+0.44%

No instruction-count regression above 1%.

_{Only instruction counts gate. Wall clock is shown for context — on identical hardware it moves 4-20% run to run.}

_{Measured by tak — instruction-counted CLI benchmarks, stored in this repository's git notes.}

Shadow comparison

Parsing mise use -g node@20 against a shadow of mise's committed spec.
Reported, not gated: the shadow grows as the derive learns to express more, so
what to watch is the ratio rather than either column.

framework	stripped binary, bytes
usage	1344728
bpaf	2493216
clap	3101656

framework	instructions, cold parse	vs usage
usage	8218	—
clap	6313214	768x
bpaf	21909171	2665x

                                              min       p01       p10    median
usage-rs: argv -> struct                     1141      1170      1179      1221  ns
clap: build tree + parse -> struct        1792822   1796017   1801295   1810452  ns
bpaf: build parser + parse -> struct      5623120   5623120   5644864   5676868  ns

usage: argv -> struct                            1212 ns      1.21 µs
clap: build tree + parse -> struct            1816139 ns   1816.14 µs
clap: parse -> struct, tree reused              72143 ns     72.14 µs
clap: build tree only                         1050259 ns   1050.26 µs

_{be4fdc1ee4f1 vs 1dd4feb5df0a · measured on the runner, not pushed to the history.}