perf: Optimize writeIntegerList

Replaces the two-pass (compute size, then write) approach with single-pass type-specific writers that reserve placeholder space for the length prefix and patch it in afterward, falling back to two-pass only for very large lists. Output bytes unchanged.

**Benefit:** cuts a full extra pass over packed integer lists on the write path — same wire format, fewer CPU cycles.

PR: hashgraph/pbj#917

Start at the writeIntegerList implementation and inspect PR hashgraph/pbj#917, which contains the proposed optimization. Verify that the single-pass writers preserve the existing wire-format bytes and that the fallback for very large lists remains covered by the change.

Written by the indexing model from the issue text.