buffrs lint exits 0 on rule violations #322

Summary

lint() collects violations, prints them, and unconditionally returns Ok(()) (command.rs:401-419):

violations
    .into_iter()
    .map(miette::Report::new)
    .for_each(|r| eprintln!("{r:?}"));

Ok(())

So lint cannot gate CI without scraping output.

Reproduction

buffrs init --lib my-protos
cd my-protos
printf 'syntax = "proto3";\npackage wrongpkg;\nmessage Foo { string bar = 1; }\n' > proto/my_protos.proto
buffrs lint; echo "exit=$?"

PackageName (https://helsing-ai.github.io/buffrs/reference/protocol-buffer-rules.html)

  × Make sure that the protobuf package name matches the buffer package name.
  ╰─▶   × package name is wrongpkg but should have my_protos prefix

   ╭─[my-protos/my_protos.proto:1:1]
   ╰────
exit=0

Note the violation goes to stderr, not stdout, so the obvious buffrs lint | grep ... workaround silently passes.

Expected

Non-zero exit when one or more violations are reported.

Actual

exit=0 regardless of violation count.

Impact

CI must do buffrs lint 2>&1 | grep -q '×' && exit 1, which is fragile and breaks on any output format change. Consumers are effectively unable to enforce lint.

Suggested fix

Return a non-zero exit status when the violation count is greater than zero. Since rules carry a severity Level already, gate on Level::Error by default and add --deny warnings / --allow warnings to make the threshold configurable.

Side note

While writing this I noticed IdentCasing is registered in the rule set but its check_entity returns Default::default() — identifier casing is never actually checked. That is a separate bug; happy to file it separately if useful.