Measure and optimize shell completion project-name latency

[codeforester]

Finding

Bash and Zsh project-name completions invoke base-wrapper --project base base_projects list when completing project arguments. Project discovery has cache support, but completion still pays process startup and wrapper overhead.

Evidence

• lib/shell/completions/basectl_completion.sh calls base-wrapper --project base base_projects list for project names.
• lib/shell/completions/basectl_completion.zsh does the same.
• This is user-facing latency on shell tab completion, so it should be measured before optimization.

Acceptance Criteria

• Measure completion latency in a representative workspace before changing behavior.
• Decide whether shell-session caching, a generated completion cache, or no change is warranted.
• If optimized, keep stale project names bounded and predictable.
• Add coverage or a lightweight smoke test for the chosen behavior.

Suggested Validation

• Document measured before/after timings in the PR.
• Relevant completion tests.
• git diff --check.