# fix(strutil): center-aligned Resize/Align overflows the requested length

## Problem

`strutil.Resize` / `strutil.Align` (and their utf8 variants) with
`comdef.Center` alignment produce strings longer than the requested length for
almost every non-empty input:

```go
strutil.Resize("ab", 6, comdef.Center)   // "  ab   " (len 7), want "  ab  " (len 6)
strutil.Resize("ab", 4, comdef.Center)   // " ab  "   (len 5), want " ab "   (len 4)
strutil.Resize("abc", 7, comdef.Center)  // "  abc   "(len 8), want "  abc  "(len 7)
```

The extra space that centers an odd amount of padding is added unconditionally
whenever the input is non-empty, not only when the padding is actually odd.

## Root cause

`strutil/padding.go`, `resize`:

```go
if align == comdef.Center {
    padLn := (wantLen - sLen) / 2
    if diff = wantLen - padLn*2; diff > 0 {  // wrong: missing "- sLen"
        s += " "
    }
    ...
}
```

The condition is meant to detect an odd amount of total padding
(`wantLen - sLen`). Simplifying the buggy expression:

```
wantLen - 2*padLn = wantLen - 2*floor((wantLen-sLen)/2)
                  = sLen + ((wantLen - sLen) mod 2)
```

so `diff > 0` is true whenever `sLen > 0`, regardless of the padding parity.
The result is one space too long for every non-empty string with even padding.

The existing `TestResize("title", 20, PosMiddle)` case did not catch this
because its padding (15) is odd, where the extra space is legitimately needed.

## Fix

Compare against the actual leftover padding by subtracting `sLen`:

```go
if diff = wantLen - sLen - padLn*2; diff > 0 {
    s += " "
}
```

This adds the extra space only when `(wantLen - sLen)` is odd, keeping the odd
case (e.g. the "title" test) unchanged.

## Test

Added `TestResize_center_length` in `strutil/padding_test.go` asserting both the
exact result and `len(result) == wantLen` for even- and odd-padding cases.

- Gate: `go test ./strutil/ -run TestResize` -> ok (PASS)
- `go vet ./strutil/` -> exit 0
- `go build ./...` -> exit 0

Counterfactual: reverting the change makes the new test fail with
`actual: int(7)` / `actual: "  ab   "` vs the expected length 6 / `"  ab  "`.

Note: `TestDateSNv3/dateLen4_base62` in `strutil/gensn_test.go` is flaky and
already fails on the unmodified base commit (it asserts uniqueness of randomly
generated serial numbers); it is unrelated to this change.