Added util.ByteSizeNoLimit: a byte size of -1 means no limit, and any other negative size is rejected

ParseByteSize accepts "-1" or "-1B" as ByteSizeNoLimit, and String gives it
back as "-1". A setting for which no limit is unsafe can reject it.

internal/util/bytesize.go

@@ -4,15 +4,21 @@ import (
 	"fmt"
 	"math"
 	"strconv"
+	"strings"
 
 	"github.com/dustin/go-humanize"
 )
 
 // ByteSize is a size in bytes, configured as a string such as "32KiB", "1MB"
 // or "1.5 GB" (see ParseByteSize). KB, MB, ... are powers of 1000 and KiB,
-// MiB, ... powers of 1024; a bare number is bytes.
+// MiB, ... powers of 1024; a bare number is bytes. ByteSizeNoLimit (-1) means
+// no limit; no other size is negative.
 type ByteSize int64
 
+// ByteSizeNoLimit is the size that means no limit, written "-1". A setting
+// for which no limit is unsafe rejects it.
+const ByteSizeNoLimit ByteSize = -1
+
 // byteUnits are the units String may use, largest first.
 var byteUnits = []struct {
 	name string
@@ -27,8 +33,16 @@ var byteUnits = []struct {
 }
 
 // ParseByteSize parses a size such as "32KiB", "1 MB" or "4096". Unit names
-// are case-insensitive.
+// are case-insensitive. "-1" (or "-1B") is ByteSizeNoLimit; any other
+// negative size is an error.
 func ParseByteSize(s string) (ByteSize, error) {
+	trimmed := strings.TrimSpace(s)
+	if trimmed == "-1" || strings.EqualFold(trimmed, "-1B") {
+		return ByteSizeNoLimit, nil
+	}
+	if strings.HasPrefix(trimmed, "-") {
+		return 0, fmt.Errorf("byte size %q is negative; only -1 (no limit) is allowed", s)
+	}
 	n, err := humanize.ParseBytes(s)
 	if err != nil {
 		return 0, err
@@ -41,7 +55,11 @@ func ParseByteSize(s string) (ByteSize, error) {
 
 // String formats b exactly, in the largest unit that divides it (e.g. "32KiB",
 // "1MB"), or in bytes ("1500B"), so it parses back to the same value.
+// ByteSizeNoLimit is "-1".
 func (b ByteSize) String() string {
+	if b == ByteSizeNoLimit {
+		return "-1"
+	}
 	if b != 0 {
 		for _, u := range byteUnits {
 			if int64(b)%u.size == 0 {