fix(formatter): classify unsupported feature errors

[overtrue]

Summary

This change closes a formatter regression in the phase-3 structured error work. The formatter already inferred unsupported_feature for messages containing unsupported or does not support, but one of the new command error strings uses the phrase not yet supported instead. That path fell through to general_error, which meant JSON error output lost the expected error type and recovery hint.

The fix keeps scope narrow. It adds focused regression tests for the message-based conflict and unsupported_feature inference branches, then extends the unsupported-feature classifier to recognize not yet supported messages. The new unsupported test failed before this change and now passes with the classifier update.

Validation

I attempted to run make pre-commit, but this repository does not define that target or a Makefile/justfile. I ran the equivalent required checks from AGENTS.md instead:

• cargo fmt --all --check
• cargo clippy --workspace -- -D warnings
• cargo test --workspace

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 17ed2ae5f7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Avoid attaching --force hint to generic "not yet supported" errors

Including "not yet supported" in the unsupported-feature matcher makes Formatter::error emit UNSUPPORTED_SUGGESTION for messages that do not have a force path, e.g. diff/mirror local-path rejections and cross-alias cp/mv (crates/cli/src/commands/diff.rs, mirror.rs, cp.rs, mv.rs), whose argument structs have no --force flag. In JSON mode this now returns a misleading recovery action (Retry with --force...) that users and clients cannot apply, whereas previously these cases had no incorrect hint.

Useful? React with 👍 / 👎.