[fallintoplace]

What

• Copy unknown length-delimited fields into their final allocation.

Why

• The decoder currently zeroes the buffer, then overwrites the same bytes.

Implementation

• Copy contiguous payloads once and copy fragmented Buf chunks into the same allocation.
• Add edge-case coverage and an opt-in Criterion benchmark.

Benchmark

• Criterion, 40 samples: 4 KiB decode went from 135 ns to 100 ns contiguous and 134 ns to 95 ns split. 32 B contiguous stayed near 36 ns; split was 3.7% slower.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}