chore: replace number-allocator with inline Set-based implementation

[roli-lpci]

Summary

Replaces the number-allocator package (which depends on js-sdsl and debug) with an 80-line inline implementation using a Set and a low-water-mark cursor.

This addresses item 1.1 of the v6 dependency cleanup roadmap (#2038).

• Removes 3 transitive dependencies (number-allocator, js-sdsl, debug)
• All existing tests pass (123 integration + 10 unit)
• 19 new dedicated unit tests for the inline NumberAllocator
• Browser-compatible (Set is ES2015)

Changes

File	Change
src/lib/number-allocator.ts	New — inline Set-based allocator (~80 lines)
src/lib/unique-message-id-provider.ts	Import path change, removed unnecessary as boolean cast
src/lib/topic-alias-send.ts	Import path change
package.json	Removed number-allocator dependency
package-lock.json	Regenerated
test/node/number-allocator.ts	New — 19 dedicated unit tests

Implementation

The inline NumberAllocator tracks used numbers in a Set<number> and maintains a lowWaterMark cursor that avoids full scans in the common sequential-allocation case. For MQTT's 16-bit ID range (1–65535), this is more than sufficient.

API is identical to the original: alloc(), use(), free(), clear(), firstVacant().

Test plan

• Existing message-id-provider tests pass (10/10)
• New NumberAllocator unit tests pass (19/19)
• Integration tests with UniqueMessageIdProvider pass (123/123)
• TypeScript compiles clean
• Browser bundle builds successfully

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.68%. Comparing base (280a311) to head (61a3e66).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2041      +/-   ##
==========================================
+ Coverage   83.41%   83.68%   +0.26%     
==========================================
  Files          25       26       +1     
  Lines        1586     1612      +26     
  Branches      367      372       +5     
==========================================
+ Hits         1323     1349      +26     
  Misses        230      230              
  Partials       33       33              

Flag	Coverage Δ
unittests	83.68% <100.00%> (+0.26%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[robertsLando]

LGTM

[Copilot]

Pull request overview

This PR removes the external number-allocator dependency by introducing an internal NumberAllocator implementation and updating internal call sites to use it, supporting the v6 dependency cleanup roadmap (#2038).

Changes:

• Added src/lib/number-allocator.ts (Set + low-water-mark allocator) and new unit tests.
• Updated internal imports to reference the new inline allocator.
• Removed number-allocator from package.json and regenerated package-lock.json.

Reviewed changes

Copilot reviewed 5 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/lib/number-allocator.ts	New inline allocator implementation to replace the number-allocator package.
src/lib/unique-message-id-provider.ts	Switched to local allocator import; removed unnecessary boolean cast.
src/lib/topic-alias-send.ts	Switched to local allocator import.
test/node/number-allocator.ts	Added dedicated unit tests for the new allocator behavior.
package.json	Removed number-allocator dependency.
package-lock.json	Lockfile updated to reflect dependency removal (plus additional upgrades).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

use() does not update lowWaterMark, so the cursor can point at a known-occupied value after calling use(lowWaterMark) (or after marking many values via use()), causing alloc()/firstVacant() to rescan already-used IDs repeatedly. Consider advancing lowWaterMark when num <= lowWaterMark (e.g., increment until a vacant slot) to maintain the intended invariant and keep the sequential-allocation optimization effective.

Suggested change

	this.used.add(num)
	this.used.add(num)
	// Maintain the invariant that lowWaterMark is the lowest number that might be vacant.
	// If we just used a number at or below lowWaterMark, advance lowWaterMark past any
	// consecutively used numbers.
	if (num <= this.lowWaterMark) {
	while (this.lowWaterMark <= this.max && this.used.has(this.lowWaterMark)) {
	this.lowWaterMark++
	}
	}

[robertsLando]

I think this is a legit change