fix(repl): print every line of a usage error at the prompt

[devin-ai-integration]

What and why

The terminal prompt printed only the first line of a multi-line usage error. A repository or notebook %load, %repo, %projects or %publish given wrong arguments answers with a *repl.UsageError whose Lines hold the usage line and the line saying what was wrong; UsageError.Error() is Lines[0] alone, and repl.Loop printed "error: " + merr.Error(). So %load --id x showed the usage and dropped --id is not an option of %load, while the Jupyter kernel, which reports execError("UsageError", usage.Error(), usage.Lines), showed both.

Two conventions coexist in the meta handlers. %render, %viz and %render-document return their usage lines as output (metaRender: return usage, false, nil), so the loop already printed them whole; *UsageError there is what the typed APIs (Session.Render, Viz, RenderDocument) return to callers like the kernel. The repository and notebook commands instead return *UsageError as the meta error (usageError(...) in repository.go, parseLoadArgs in notebook.go), which is what the kernel's %projects now → UsageError test and docs/reference/jupyter-kernel.md ("a % command given wrong arguments is a UsageError whose traceback is the command's usage") rely on. Moving those handlers to the output convention would turn the kernel's UsageError into plain stdout, so the fix is on the terminal side: repl.Loop now prints every line of a *UsageError through metaErrorLines — the first under the error: prefix, as before and as the kernel's value is; the rest as they are, as the kernel's traceback shows them. Any other meta error prints as it did. UsageError.Error() is unchanged, so the kernel's output and its tests are untouched.

No other consumer loses lines: cmd/sysml reaches RunMeta only through repl.Loop, and the referee's %eval in path never produces a UsageError.

How it was verified

New TestLoopPrintsEveryLineOfAUsageError in internal/frontend/repl/repository_test.go drives %load --id x, %load --id=a --id=b and %projects now through repl.Loop and pins the exact lines. No existing test pinned the one-line terminal output; the kernel tests (internal/frontend/jupyter/{repository,replengine}_test.go) still pass unchanged.

%load --id x at the prompt, before:

sysml> %load --id x
error: usage: %load [--id=<project id>] [--name=<name>] [--branch=<branch name or id>] [<name>]

after:

sysml> %load --id x
error: usage: %load [--id=<project id>] [--name=<name>] [--branch=<branch name or id>] [<name>]
--id is not an option of %load

Gates run locally: gofmt -l . (empty), go build ./..., go vet ./..., make lint, go test -race -count=1 ./internal/frontend/repl/... ./internal/frontend/jupyter/..., go test ./..., python3 scripts/changelog.py check, python3 scripts/check-doc-ids.py.

Checklist

• make test and make lint pass locally
• Tests added or updated for the change
• Documentation extended where it already covers the surface (see CONTRIBUTING.md) — the reference documents the kernel's UsageError; the prompt's output needed no new text
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (no gate count moved)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

🔍 Devin Review: 2 flags

Not posted on this PR by your GitHub settings — view them in Devin Review. (Configure)