docs(CompressedIntSet): say four bytes of *payload* for a fresh-chunk range add

Addresses round 7 of the Copilot review on #337 (one point, on two doc surfaces).

"A range in a fresh chunk is four bytes, whatever its width" described the run
pair's payload, not its heap cost — the pair lives in a ushort[], which carries an
array header. Both surfaces now say "four bytes of payload" and note the header,
matching how MemoryUsageInBytes already documents its own exclusion. The claim the
sentence exists to make — that the cost is independent of the range's width — is
unchanged and still true.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: marius-bughiu

docs/api/collections.md

@@ -2224,7 +2224,8 @@ CompressedIntSet(IEnumerable<int> source)
 - `bool TryAdd(int item)` — `true` if added, `false` if already present.
 - `long AddRange(int start, int endInclusive)` — adds every value in the inclusive range and
   returns how many were **new**. A range landing in a chunk the set does not yet touch is stored as
-  a **single run pair — four bytes, whatever the range's width** — so this is the cheap way to
+  a **single run pair — four bytes of payload, whatever the range's width** (plus the array header,
+  which `MemoryUsageInBytes` excludes too) — so this is the cheap way to
   build a clustered set. Throws `ArgumentOutOfRangeException` if `endInclusive < start`.
 - `bool Contains(int item)` — `O(1)` in a bitmap chunk, `O(log n)` in an array or run chunk, and a
   single comparison against the chunk index when nothing covers the value.

src/Celerity/Collections/CompressedIntSet.cs

@@ -243,7 +243,8 @@ public bool TryAdd(int item)
     /// <returns>The number of values the call actually added (values already present are not counted).</returns>
     /// <remarks>
     /// A range that lands in a chunk the set does not yet touch is stored as a single run pair —
-    /// four bytes, whatever the range's width — which is why this is the cheap way to build a
+    /// four bytes of payload, whatever the range's width (plus the array header, which
+    /// <see cref="MemoryUsageInBytes"/> excludes too) — which is why this is the cheap way to build a
     /// clustered set. A range overlapping an existing chunk merges into it and the chunk is left in
     /// its natural form; call <see cref="Optimize"/> afterwards to re-compress those.
     /// </remarks>