perf: optimize jump command search to single-pass algorithm

[kexi]

Context

From PR #324 review feedback (non-blocking suggestion).

Problem

The current vibe jump search logic performs multiple passes over the worktree list:

1. Exact match (case-sensitive)
2. Exact match (case-insensitive)
3. Word boundary match (case-sensitive)
4. Word boundary match (case-insensitive)
5. Substring match (case-sensitive)
6. Substring match (case-insensitive)

Proposal

Optimize with a single pass that categorizes matches into buckets, then selects the highest-priority non-empty bucket.

Priority

Low — typical repositories have <10 worktrees, making the current multi-pass approach negligible in practice. This becomes relevant only for repositories with hundreds of worktrees.

Related

• PR feat: add jump command for worktree navigation #324