Summary

Makes all 16 previously-failing tests pass (now 60 pass / 0 fail) without modifying any test files or adding dependencies. Only source files under src/ were changed.

Changes by file

• src/string-utils.ts — wordCount now trims and splits on \s+ so consecutive spaces don't produce empty tokens; implemented truncate (returns unchanged within limit, breaks at word boundary within the maxLength - ellipsis budget, ellipsis counts toward the limit).
• src/task-manager.ts — implemented remove (returns false for unknown id), update (title/priority, returns false for unknown id), and sortBy for priority (high → medium → low) and createdAt (oldest first).
• src/validator.ts — isEmail accepts long TLDs (e.g. .museum); isUrl correctly handles URLs with a port.
• src/calculator.ts — divide throws on division by zero.
• src/date-utils.ts — fixed formatRelative off-by-one: tier selection now branches on the rounded magnitude so boundary cases (23.5h, 59.5min) roll up correctly instead of rendering "24 hours"/"60 minutes".

Verification

• bun test: 60 pass, 0 fail (70 assertions).
• No files under test/ were modified; no dependencies added.
• Reviewed by the review subagent, which flagged one boundary defect in the initial formatRelative fix — that was corrected before this PR.

Assumptions

• The tests are the source of truth for intended API/edge-case behaviour.
• formatRelative tier rounding was corrected to be internally consistent, a small behavioural improvement beyond the minimum required by the failing test.