Paths are joined with client side separators when targeting remotes (#547)

Currently, filepath.Join is used to join path fragments that are supposed
to be created on remote storage backends. This could theoretically cause
problems when the separators used by the client and the remotes do not match.
It's unlikely this causes problems right now, but it's definitely better to
rectify it before it causes further confusion.

This was raised in #541

internal/storage/azure/azure.go

@@ -9,6 +9,7 @@ import (
 	"errors"
 	"fmt"
 	"os"
+	"path"
 	"path/filepath"
 	"strings"
 	"sync"
@@ -128,7 +129,7 @@ func (b *azureBlobStorage) Copy(file string) error {
 	_, err = b.client.UploadStream(
 		context.Background(),
 		b.containerName,
-		filepath.Join(b.DestinationPath, filepath.Base(file)),
+		path.Join(b.DestinationPath, filepath.Base(file)),
 		fileReader,
 		b.uploadStreamOptions,
 	)
@@ -141,7 +142,7 @@ func (b *azureBlobStorage) Copy(file string) error {
 // Prune rotates away backups according to the configuration and provided
 // deadline for the Azure Blob storage backend.
 func (b *azureBlobStorage) Prune(deadline time.Time, pruningPrefix string) (*storage.PruneStats, error) {
-	lookupPrefix := filepath.Join(b.DestinationPath, pruningPrefix)
+	lookupPrefix := path.Join(b.DestinationPath, pruningPrefix)
 	pager := b.client.NewListBlobsFlatPager(b.containerName, &container.ListBlobsFlatOptions{
 		Prefix: &lookupPrefix,
 	})