Rebuild: chunk format v2 — nyuu and hex message-ids as bits

nyuu's 12 capital+small letter pairs in 15 bytes (10 bits a pair), its
timestamp as a delta within the file, the suffix once per file;
ngPost-style 32 hex digits in 16 bytes; other ids as zstd text as before.
Recent boneless: 25.2 → 20.3 B per segment (and 25.0 → 19.5 on the next
range); a 2013 ath range stays at 17.9 (JBinUp ids are text). Every chunk
of three real ranges decodes to itself (edge:bench --verify). Encoding is
no slower than v1 (less text for zstd). Version 1 chunks stay readable.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

commands/edge_bench.ts

@@ -7,6 +7,7 @@ import env from '#start/env'
 import { XoverClient } from '#lib/nntp/index'
 import { RangeIndexer } from '#lib/edge/index'
 import { DEFAULT_RULES, type Rules } from '#lib/rules/index'
+import { decodeChunk, encodeChunk } from '#lib/chunks/index'
 import db from '@adonisjs/lucid/services/db'
 import BatchStore from '#services/batch_store'
 
@@ -45,6 +46,9 @@ export default class EdgeBench extends BaseCommand {
   @flags.number({ description: 'Chunks without a new part before an open file is given up', default: 5 })
   declare idle: number
 
+  @flags.boolean({ description: 'Decode every release chunk and encode it again: the same bytes' })
+  declare verify: boolean
+
   @flags.boolean({ description: 'Also store the batches (BatchStore, the configured database), timed' })
   declare store: boolean
 
@@ -137,6 +141,12 @@ export default class EdgeBench extends BaseCommand {
       out.releases += batch.releases.length
       batches++
       for (const r of batch.releases) {
+        if (this.verify) {
+          const decoded = decodeChunk(r.payload)
+          if (!encodeChunk(decoded).equals(r.payload) || decoded.reduce((n, f) => n + f.segments.length, 0) !== r.segments) {
+            throw new Error(`Release ${r.key}: its chunk doesn't decode to itself`)
+          }
+        }
         keys.add(r.key)
         out.files += r.files
         out.segments += r.segments