perf(transport): pool *bytes.Buffer + MaxBytesReader for webhook decode

Replace the hand-rolled make([]byte, 0, 1024) + make([]byte, 4096) read loop in WebhookServer.ServeHTTP with a sync.Pool-backed bytes.Buffer drained via ReadFrom, fronted by http.MaxBytesReader for the 1 MiB body cap.

putWebhookBuf caps Cap() at 256 KiB before returning to the pool so a rare oversized update (max body is 1 MiB) doesn't permanently bloat the pool.

Bench delta on Webhook_ServeHTTP: 2564ns -> 2020ns (-21%), 12707B -> 7648B (-40%), 24 -> 23 allocs. The big byte saving is the 4 KiB tmp buffer + 1 KiB initial buf cap, replaced by one reused buffer across requests. The remaining alloc count is dominated by codec.Unmarshal decoding Update's pointer fields (*string, *int64), which is downstream of this change.

.benchstats/after_step5.txt

@@ -0,0 +1,30 @@
+After: static-headers + bool-fast-path + lazy-Values + typed-fields + resp-buffer-pool + webhook-pool
+goos: darwin
+goarch: arm64
+cpu: Apple M4 Max
+
+pkg: github.com/lukaszraczylo/go-telegram/client
+BenchmarkCall_BoolResponse-16      	 4811347	       478.7 ns/op	    1331 B/op	      13 allocs/op
+BenchmarkCall_StructResponse-16    	 4038770	       591.6 ns/op	    1462 B/op	      15 allocs/op
+BenchmarkEncodeJSONBody-16         	47025052	        51.30 ns/op	      96 B/op	       2 allocs/op
+BenchmarkDecodeResult_Bool-16      	853161562	         2.824 ns/op	       0 B/op	       0 allocs/op
+BenchmarkDecodeResult_Struct-16    	26811634	        88.80 ns/op	     144 B/op	       2 allocs/op
+
+pkg: github.com/lukaszraczylo/go-telegram/transport
+BenchmarkWebhook_ServeHTTP-16    	 1204390	      2020 ns/op	    7648 B/op	      23 allocs/op
+
+pkg: github.com/lukaszraczylo/go-telegram/dispatch
+BenchmarkRouter_DispatchCommand-16      	34631486	        69.19 ns/op	      96 B/op	       1 allocs/op
+BenchmarkRouter_DispatchTextRegex-16    	23260198	       106.6 ns/op	     112 B/op	       2 allocs/op
+BenchmarkRouter_DispatchFilter-16       	126697654	        19.03 ns/op	      96 B/op	       1 allocs/op
+BenchmarkRouter_NewContext-16           	1000000000	         1.600 ns/op	       0 B/op	       0 allocs/op
+BenchmarkExtractCommand-16              	27345622	        87.25 ns/op	       0 B/op	       0 allocs/op
+
+Cumulative deltas vs baseline:
+  Call_BoolResponse:    634ns / 18 allocs / 1957B   ->  479ns / 13 / 1331B   (-24% / -5  / -626B)
+  Call_StructResponse:  665ns / 18 allocs / 2005B   ->  592ns / 15 / 1462B   (-11% / -3  / -543B)
+  DecodeResult_Bool:     50ns /  2 allocs /   80B   -> 2.8ns /  0 /    0B
+  Webhook_ServeHTTP:   2564ns / 24 allocs / 12707B  -> 2020ns / 23 / 7648B   (-21% / -1  / -5059B)
+  DispatchCommand:      153ns /  5 allocs /  416B   ->   69ns /  1 /   96B   (-55% / -4  / -320B)
+  DispatchTextRegex:    181ns /  5 allocs /  428B   ->  107ns /  2 /  112B   (-41% / -3  / -316B)
+  DispatchFilter:        32ns /  2 allocs /   96B   ->   19ns /  1 /   96B   (-41% / -1)

transport/webhook.go

@@ -6,6 +6,7 @@
 package transport
 
 import (
+	"bytes"
 	"context"
 	"crypto/subtle"
 	"errors"
@@ -18,6 +19,24 @@ import (
 	"github.com/lukaszraczylo/go-telegram/client"
 )
 
+// webhookBufPool reuses *bytes.Buffer for incoming webhook bodies.
+// Webhook payloads are typically a single Telegram Update (commonly
+// 1-8 KiB), so a buffer that has grown once will satisfy most
+// subsequent requests with no additional allocation.
+var webhookBufPool = sync.Pool{New: func() any { return new(bytes.Buffer) }}
+
+// maxWebhookBufCap caps the buffer size returned to webhookBufPool so
+// a rare oversized update doesn't permanently bloat the pool. Buffers
+// larger than this are dropped on the floor.
+const maxWebhookBufCap = 256 * 1024
+
+func putWebhookBuf(buf *bytes.Buffer) {
+	if buf.Cap() > maxWebhookBufCap {
+		return
+	}
+	webhookBufPool.Put(buf)
+}
+
 // WebhookServer implements Updater by exposing an http.Handler that
 // receives updates from Telegram. It can be mounted on the user's own
 // HTTP server (via ServeHTTP) or run standalone (via ListenAndServe).
@@ -108,28 +127,27 @@ func (w *WebhookServer) ServeHTTP(rw http.ResponseWriter, r *http.Request) {
 	}
 	w.handlers.Add(1)
 	defer w.handlers.Done()
+
+	const maxBody = 1 << 20 // 1 MiB cap on body
+	r.Body = http.MaxBytesReader(rw, r.Body, maxBody)
 	defer func() { _ = r.Body.Close() }()
 
-	const max = 1 << 20 // 1 MiB cap on body
-	buf := make([]byte, 0, 1024)
-	tmp := make([]byte, 4096)
-	for {
-		n, err := r.Body.Read(tmp)
-		if n > 0 {
-			buf = append(buf, tmp[:n]...)
-			if len(buf) > max {
-				rw.WriteHeader(http.StatusRequestEntityTooLarge)
-				return
-			}
-		}
-		if errors.Is(err, http.ErrBodyReadAfterClose) || err != nil {
-			break
+	buf := webhookBufPool.Get().(*bytes.Buffer)
+	buf.Reset()
+	defer putWebhookBuf(buf)
+	if _, err := buf.ReadFrom(r.Body); err != nil {
+		var maxErr *http.MaxBytesError
+		if errors.As(err, &maxErr) {
+			rw.WriteHeader(http.StatusRequestEntityTooLarge)
+			return
 		}
+		rw.WriteHeader(http.StatusBadRequest)
+		return
 	}
 
 	var u api.Update
 	codec := w.Bot.Codec()
-	if err := codec.Unmarshal(buf, &u); err != nil {
+	if err := codec.Unmarshal(buf.Bytes(), &u); err != nil {
 		rw.WriteHeader(http.StatusBadRequest)
 		return
 	}