fix(docs): unblock markdown --append on Thai/CJK/emoji content

[ninyawee]

Summary

Fixes #559 — gog docs write --markdown --append hung indefinitely (CPU-bound infinite loop) on any markdown content ending in a non-ASCII rune. Thai content reliably triggered it because Thai paragraphs/headings naturally end with multi-byte runes; the same bug applies to CJK and emoji content.

The hang was in nextRune (internal/cmd/docs_markdown.go). For a string containing exactly one multi-byte rune, the previous range-based implementation:

for i, r := range s {
    if i > 0 { return s[:i], i }   // never fires (only one iteration)
    if len(s) == 1 { return s, 1 } // false: len() is bytes, not runes
    _ = r
}
return "", 0   // ← falls through here

returned size 0, and the caller in ParseInlineFormatting advanced currentByte += 0 forever inside its for currentByte < len(text) loop.

--markdown --replace is unaffected because that path uploads to Drive's server-side markdown converter (Media(strings.NewReader(cleaned), ContentType(mimeTextMarkdown)) in docs_edit.go). Only --append and the find-replace/sed paths invoke the local parser, so this issue affects:

• docs write --markdown --append
• docs find-replace --format markdown (when replacement text ends in a multi-byte rune)
• docs sed family commands using markdown replacement content

Fix

Replace the hand-rolled rune iteration with utf8.DecodeRuneInString. It handles every case correctly — empty input, ASCII, multi-byte runes, and invalid UTF-8 (RuneError with its byte size).

Test plan

• Added unit tests for nextRune covering the previously-broken paths (single Thai rune, single emoji, empty input, mixed strings).
• Added timeout-guarded tests for ParseInlineFormatting on Thai headings, FAQ-style mixed text, bold/italic/code mixed with Thai, and trailing emoji. Each call is wrapped in a 2s timeout so any future infinite-loop regression fails fast in CI rather than stalling it.
• Added an end-to-end test TestMarkdownToDocsRequests_ThaiAppend exercising the exact path used by docs write --markdown --append — ParseMarkdown → MarkdownToDocsRequests at a non-zero base index — on a multi-element Thai sample (heading + paragraph + bullet list + blockquote).
• Verified the new tests fail on main HEAD (each times out and the goroutine stack confirms it's stuck inside ParseInlineFormatting at line 434) and pass with the fix.
• make lint clean. Full go test ./internal/cmd/... passes apart from a pre-existing failure in TestRequireAccount_Missing that is unrelated to this change (also fails on stock main).
• Verified end-to-end against the live Google Docs API:
  • The repro case (--markdown --append on a single-line Thai heading) used to time out at 30s; now completes in ~1.4s.
  • A 14 KB Thai markdown doc with 136 generated batch-update requests now appends in ~1.7s.

Notes

• Issue docs write --markdown --append hangs forever on Thai content (infinite loop in nextRune) #559 was filed by me from the same investigation; closing it on merge.
• I am running gogcli via mise (github:steipete/gogcli@0.14.0); the bug reproduces identically on current main (56755e9) which is what this branch targets.