fix: stop TUI truncation helpers from splitting multi-byte UTF-8 runes

[STiFLeR7]

Summary

truncStr, firstN, lastN (internal/render/helpers.go), ptrunc (proxy.go), and truncateShellLine (terminal.go) all slice on raw byte offsets. When a cut lands mid-rune, the output is invalid UTF-8 and the terminal shows �. Same root cause cuts CJK/emoji lines short, since those runes spend 2-4 bytes but only 1-2 display columns of the truncation budget.

Fix

Switch all five to github.com/charmbracelet/x/ansi (Truncate/TruncateLeft), already a direct dependency and already the established pattern for this exact problem in internal/app (view.go). Both are grapheme-aware and count display columns, not bytes.

• truncStr/firstN ("keep the start"): ansi.Truncate(s, n, "")
• lastN ("keep the end"): reversed via ansi.TruncateLeft(s, w-n, "") where w is the string's total display width — ansi.TruncateLeft removes N columns from the start, so keeping the last n columns means removing w - n.
• ptrunc/truncateShellLine (truncate-with-ellipsis): ansi.Truncate already reserves the tail string's width internally, so this is a direct swap of the same max/maxLineLength budget — no manual -3 arithmetic needed anymore.

Fixes #1152.

Testing

• Added helpers_test.go with the issue's own repro (TestTruncSplitsRunes, asserting valid UTF-8 out of all five paths) plus per-function behavior tests (TestFirstNTruncStrKeepStart, TestLastNKeepsEnd, TestPtruncAppendsEllipsisWithinBudget).
• Verified red-before-green: reverted the fix (kept the new tests), confirmed the exact reported mojibake (e.g. firstN("中文字ABC", 4) → "中\xe6", invalid UTF-8); restored the fix, confirmed valid output.
• go build ./..., go vet ./..., and the full existing internal/render test suite all pass with the fix applied.
• gofmt -l reports no formatting issues.

[greptile-apps]

Greptile Summary

The PR replaces byte-offset TUI truncation with grapheme-aware, display-column-based ANSI helpers to prevent invalid UTF-8 and mojibake.

• Updates start-preserving, end-preserving, proxy, and shell-output truncation.
• Adds regression coverage for CJK text, UTF-8 validity, suffix retention, and ellipsis behavior.

Confidence Score: 5/5

The PR appears safe to merge, with no concrete blocking or independently actionable non-blocking issues identified.

The changed helpers consistently use display-column-aware truncation, preserve valid UTF-8, retain existing caller budgets, and are covered by targeted regression tests.

Important Files Changed

Filename	Overview
strix/interface/tui/internal/render/helpers.go	Replaces byte slicing in shared truncation helpers with display-width-aware truncation while preserving start/end semantics.
strix/interface/tui/internal/render/helpers_test.go	Adds focused regression tests for valid UTF-8, CJK display widths, suffix truncation, and ellipsis budgets.
strix/interface/tui/internal/render/proxy.go	Uses ANSI-aware truncation for bounded proxy-rendering values without changing reachable caller budgets.
strix/interface/tui/internal/render/terminal.go	Makes shell-line truncation grapheme-safe while retaining the existing 200-column limit and ellipsis behavior.

_{Reviews (1): Last reviewed commit: "fix: stop TUI truncation helpers from sp..." | Re-trigger Greptile}