feat(cd): add 'wt cd' to switch between existing worktrees

[timvw]

Closes #103.

Why

wt co already has a fuzzy picker and auto-cd, but its list comes from getAvailableBranches() — every local and remote branch — and picking a branch without a worktree creates one. The request in #103 is narrower: jump between the worktrees that already exist, which is what cd $(wt ls | fzf | cut -d' ' -f1) does today.

What

wt cd [branch] (alias wt sw):

• Without arguments: fuzzy picker over existing worktrees only, including the main checkout. The current worktree is labelled (current); detached worktrees fall back to their directory name.
• With a branch: navigates directly, or fails with no worktree found for branch: X / Use 'wt checkout X' to create one.
• Never creates a worktree or a branch.
• --format json emits the usual envelope with path / navigate_to, and requires an explicit branch argument (same rule as checkout and remove).

Tab completion for cd/sw added for bash, zsh, fish and PowerShell, completing existing worktree branches (unlike remove, the main checkout is included). Docs updated in README, llms.txt, AGENTS.md, the wt examples catalog, and the Claude Code skill.

One implementation note

wt cd prints ✓ Switched to worktree: <path> before the cd marker. That is not just cosmetic: the shell wrapper extracts the path with grep '^wt navigating to: ', and script(1) can prefix the first output line with a control character (^D), which makes the anchored grep miss it and silently skips auto-cd. Every other navigating command happens to print a status line first, so this was never hit before. The e2e suite reproduced it immediately.

Testing

• New e2e/scenarios/cd.yaml: navigate to existing worktree, sw alias, refusal to create, return to the main checkout, JSON behaviour.
• New unit tests for target building ((current) marking, detached fallback, bare worktree skipped) and path containment.
• New pty test TestInteractiveCdWithoutArgs asserting the picker offers a branch with a worktree and not one without.
• go test ./..., go vet ./..., go run e2e/run.go (310 passed, 0 failed).

TestBashTabCompletionForCheckoutBranch and TestBashTabCompletionForCommands fail on my macOS box (bash 3.2), but they fail identically on main — unrelated to this change.

Unrelated harness bug spotted

e2e/run.go never resets __exit_code between steps, and uses ${__exit_code:-0}, so any zero-exit step following a non-zero-exit step inherits the old 1 and fails. I worked around it by splitting a scenario rather than touching shared infra here — happy to send a separate one-line fix.

[codecov]

Codecov Report

❌ Patch coverage is 41.11111% with 53 lines in your changes missing coverage. Please review.
✅ Project coverage is 36.91%. Comparing base (57d7103) to head (48f3221).

Files with missing lines	Patch %	Lines
cmd/cd.go	26.98%	45 Missing and 1 partial ⚠️
cmd/shellenv.go	73.07%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #117      +/-   ##
==========================================
+ Coverage   36.70%   36.91%   +0.20%     
==========================================
  Files          27       28       +1     
  Lines        3111     3199      +88     
==========================================
+ Hits         1142     1181      +39     
- Misses       1884     1931      +47     
- Partials       85       87       +2     

Files with missing lines	Coverage Δ
cmd/examples.go	67.74% <ø> (ø)
cmd/root.go	70.45% <100.00%> (+0.33%)	⬆️
cmd/shellenv.go	78.51% <73.07%> (-0.26%)	⬇️
cmd/cd.go	26.98% <26.98%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.