fix(cli): delete only removes tb when it is tracebloc's own alias

[LukasWodka]

Summary

tracebloc delete (offboarding) removed any sibling file named tb, guarded only by tb != exe — with no check that it was tracebloc's own alias. If a user had an unrelated tb on the same PATH dir tracebloc installed into (another tool's binary, or a symlink pointing elsewhere), offboarding deleted it — wrong-target deletion of a file tracebloc never created.

This came out of the FR review of the offboarding surface (§7.10). It's not a Bugbot finding — Bugbot's five findings on #164 are all resolved (by #165, #168, #169); this is the one the review caught that they didn't.

Fix — make delete symmetric with install

The installer already refuses to clobber a pre-existing tb (scripts/install.sh: creates it only when absent, or when readlink tb == PREFIX/tracebloc). Delete now mirrors that ownership test:

• New aliasStatus(tb, exe) (exists, ours bool) — tb is ours only when it's a symlink resolving to this binary.
• ours → remove it (unchanged behavior for a real install).
• exists but not ours (regular file, or symlink elsewhere) → left in place, with a note: "Left … in place — it isn't tracebloc's tb alias." (mirrors the installer's "already exists and isn't ours" message).
• Relative symlink targets resolve against the link's dir; an EvalSymlinks fallback covers the binary itself being a symlink (Intel Homebrew).

Tests

• TestDelete_OwnTBAlias_Removed — a real tb → tracebloc symlink is removed.
• TestDelete_ForeignTBAlias_Left — a foreign regular-file tb is not removed, is reported, and is untouched on disk. Verified to fail without the guard (unfixed removeSelf deletes it).
• The full-sequence test now lays down a real symlink so the removal path is exercised honestly.
• Full internal/... suite + go vet green.

Targets develop per convention.

🤖 Generated with Claude Code