TUI truncation helpers split multi-byte UTF-8 runes (mojibake)

[c879873067877881111]

Describe the bug

The truncation helpers in the TUI slice on byte offsets instead of rune boundaries, so when a cut lands in the middle of a multi-byte character the output is broken UTF-8 and the terminal shows a �. Same root cause also means CJK/emoji lines get cut short, since those runes eat 2-4 bytes of the budget but only take 1-2 columns.

Helpers doing this, all under strix/interface/tui/internal/render/:

• helpers.go: truncStr, firstN, lastN
• proxy.go: ptrunc (and psanitize/joinTrunc via it)
• terminal.go:87

I noticed it in the proxy view, which prints captured hosts/paths/headers/bodies — those carry UTF-8 all the time (JSON APIs, non-ASCII bodies, IDN hosts), so it's easy to hit.

To Reproduce

Helpers are unexported so easiest is a test in package render (commit 1c499c5):

func TestTruncSplitsRunes(t *testing.T) {
	for _, out := range []string{
		firstN("中文字ABC", 4),
		lastN("ABC中文字", 4),
		truncStr("中文字ABC", 4),
		ptrunc("中文字ABCDEF", 5),
	} {
		if !utf8.ValidString(out) {
			t.Errorf("invalid UTF-8: %q", out)
		}
	}
}

All four fail, e.g. firstN("中文字ABC", 4) returns "中\xe6" — the trailing byte is half a rune.

Expected behavior

Truncation shouldn't split a rune, and the width limit should count columns, not bytes.

ansi.Truncate from charmbracelet/x/ansi does exactly this and is already a dependency (used in internal/app), so switching these helpers over would fix it and match how app/ already handles width. ansi.Truncate("中文字ABC", 4, "") → "中文".

System Information:

• OS: macOS 14 arm64 (logic is platform-independent)
• Commit: 1c499c5
• Python Version: N/A, this is the Go TUI
• LLM Used: N/A, pure rendering path

Happy to send a PR switching these to ansi.Truncate with a regression test if that's the direction you want.