Use git-discovered repo root for xtask repo-level paths #531

Summary

Several Echo xtask repo-level commands derive their root from std::env::current_dir(). Replace those repo-root assumptions with the existing Git-discovered root helper (git rev-parse --show-toplevel) or an equivalent shared Rust boundary.

Problem

env::current_dir() is the invocation directory, not necessarily the repository root. Commands that read repo files, create method artifacts, check release files, or display repo-relative paths can become brittle when invoked from a subdirectory, an IDE task, an agent harness, or a nested workspace command.

Echo already has find_repo_root() in xtask/src/main.rs backed by git rev-parse --show-toplevel; the issue is that some repo-level code paths still bypass it.

Evidence

Current scan found std::env::current_dir() repo-root-style uses in xtask/src/main.rs, including method workspace commands around method_workspace, run_method_inbox, run_method_close, and run_method_pull. The same file already defines find_repo_root() using git rev-parse --show-toplevel, so this should be a consolidation rather than new discovery logic.

Desired Outcome

• Repo-level xtask commands resolve a checkout root through Git discovery instead of assuming cwd is the root.
• Cwd remains available only for behavior that intentionally means invocation directory.
• Display paths remain stable and repo-relative when commands are run from subdirectories.
• Failure outside a Git checkout is explicit and actionable.

Acceptance Criteria

• A regression test or command witness proves a representative repo-level xtask command works when invoked from a nested directory.
• xtask has one shared repo-root discovery boundary for repo-level paths.
• Existing cargo xtask commands keep their root-invocation behavior unchanged.
• Intentional crate-local test paths using CARGO_MANIFEST_DIR are left alone unless they are actually claiming repo-root semantics.