Reduce allocations when building ClassNameTries

[mcculls]

What Does This Do

Instead of allocating a BitSet on every put, each builder now has its own re-usable cache to record the jump indexes to update after we know how much content was added to the trie.

Motivation

Originally the builder was only used to create tries at build-time- since then its use has widened to include building tries at runtime. This optimization reduces allocations for that particular use-case.

Contributor Checklist

Jira ticket: [PROJ-IDENT]

[Copilot]

Pull request overview

This PR optimizes ClassNameTrie.Builder to reduce per-put() allocations by replacing the per-insertion BitSet with a reusable, builder-scoped int[] cache of jump indices that need offset updates after insertion size is known. This aligns with the builder’s expanded usage at runtime, where allocation pressure matters more.

Changes:

• Removed BitSet allocation in insertMapping and replaced it with a reusable jumpIndexes + jumpIndexCount cache.
• Added recordJumpIndex(int) helper to grow and populate the cached jump-index list.
• Simplified longJumps growth logic using Arrays.copyOf(...).

---

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

[bric3]

Nice optimization !