Drop redundant fixed in HttpSys SetDataChunkWithPinnedData

[unsafePtr]

Description

Drops the fixed (byte* ptr = bytes) block from the two SetDataChunkWithPinnedData helpers in HttpSys. The pin was redundant: the implementation captures ptr into chunk.pBuffer for HTTP.SYS to read later, so the fixed scope ends before the value is consumed. Safety always relied on the caller keeping the data alive, not the fixed block — the helper's name reflects that contract.

All 5 call sites verified

Call site	Data	Why already at a stable address
ResponseBody.cs:196, :248, :252	Helpers.ChunkTerminator / CRLF	"...\r\n..."u8 — embedded in assembly RVA section, never moved
ResponseBody.cs:275	span over handle.AddrOfPinnedObject() + offset	GCHandle.Alloc(..., Pinned), freed by FreeDataBuffers
ResponseStreamAsyncResult.cs:70, :134	Helpers.CRLF	same UTF-8 literal

UTF-8 literals ("..."u8) live in assembly metadata, not on the GC heap. The GCHandle-pinned site keeps the buffer alive for the lifetime of the chunk's use.

Behavior preservation

fixed (byte* ptr = bytes) returns null when bytes.IsEmpty, regardless of how the span was constructed. MemoryMarshal.GetReference(bytes) does not — per its documentation, for an empty span built from a non-null pointer (e.g. a zero-length ArraySegment of a GCHandle-pinned array) it returns a non-null ref. The bytes.IsEmpty ? null : ... guard preserves the original null-on-empty semantic.

Consistency with existing code

ResponseStreamAsyncResult.cs:80 already uses the equivalent pattern for an externally-pinned array — (void*)Marshal.UnsafeAddrOfPinnedArrayElement(...) with no fixed block. The two helpers were the outliers in this file. This change aligns them with that neighboring pattern.

No behavior change at runtime; pure cleanup.

[dotnet-policy-service]

Thanks for your PR, @unsafePtr. Someone from the team will get assigned to your PR shortly and we'll get it reviewed.