azureblob: fix double accounting for multipart uploads - fixes #8718

Before this change multipart uploads using OpenChunkWriter would
account for twice the space used.

This fixes the problem by adjusting the accounting delay.

backend/azureblob/azureblob.go

@@ -51,6 +51,7 @@ import (
 	"github.com/rclone/rclone/lib/env"
 	"github.com/rclone/rclone/lib/multipart"
 	"github.com/rclone/rclone/lib/pacer"
+	"github.com/rclone/rclone/lib/pool"
 	"golang.org/x/sync/errgroup"
 )
 
@@ -2670,6 +2671,13 @@ func (w *azChunkWriter) WriteChunk(ctx context.Context, chunkNumber int, reader
 		return -1, err
 	}
 
+	// Only account after the checksum reads have been done
+	if do, ok := reader.(pool.DelayAccountinger); ok {
+		// To figure out this number, do a transfer and if the accounted size is 0 or a
+		// multiple of what it should be, increase or decrease this number.
+		do.DelayAccounting(2)
+	}
+
 	// Upload the block, with MD5 for check
 	m := md5.New()
 	currentChunkSize, err := io.Copy(m, reader)