imagepull/BuildExt4: omit positional fs-size; rely on file truncation

mkfs.ext4's positional fs-size is documented in 1 KiB units (not the
filesystem's 4 KiB block size), so passing sizeBytes/4096 made
filesystems 1/4 the intended size. A 4 GiB request became a 1 GiB
ext4 in a 4 GiB file, packed to 0 free blocks — VM create then failed
with 'Could not allocate block' when patchRootOverlay tried to write
guest config.

The file is truncated to the target size before mkfs runs; without
the positional arg, mkfs uses the whole device.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

internal/imagepull/ext4.go

@@ -5,7 +5,6 @@ import (
 	"errors"
 	"fmt"
 	"os"
-	"strconv"
 
 	"banger/internal/system"
 )
@@ -53,6 +52,11 @@ func BuildExt4(ctx context.Context, runner system.CommandRunner, srcDir, outFile
 		return err
 	}
 
+	// mkfs.ext4's positional `fs-size` is documented in 1 KiB units
+	// (NOT the filesystem's 4 KiB block size), so dividing by 4096
+	// produces a filesystem 1/4 the intended size. Omit the positional
+	// entirely — the file was truncated to sizeBytes above, and mkfs
+	// with no fs-size arg uses the whole device.
 	out, runErr := runner.Run(ctx, "mkfs.ext4",
 		"-F",
 		"-q",
@@ -60,7 +64,6 @@ func BuildExt4(ctx context.Context, runner system.CommandRunner, srcDir, outFile
 		"-L", "banger-rootfs",
 		"-E", "root_owner=0:0",
 		outFile,
-		strconv.FormatInt(sizeBytes/4096, 10), // size in 4 KiB blocks
 	)
 	if runErr != nil {
 		_ = os.Remove(outFile)