fix(opencode): match canonically-equivalent Unicode in apply_patch - #32216

Summary

apply_patch fails with Failed to find expected lines in <file> when the file on disk and the model's context lines are canonically-equivalent Unicode but different byte sequences (NFC vs NFD). None of the 4 existing seekSequence passes handle combining-character equivalence, so any NFD content on disk is permanently unpatchable — the model re-reads, retries, fails, and burns tokens in a loop.

Root cause

seekSequence (packages/opencode/src/patch/index.ts) had 4 passes:

1. exact
2. rstrip
3. trim
4. normalizeUnicode + trim — but normalizeUnicode only maps smart quotes / dashes / ellipsis / NBSP; it never normalizes combining characters.

A file with decomposed й (и U+0438 + U+0306) can never match a patch containing precomposed й (U+0439), even though the strings are canonically equivalent. Models essentially always emit NFC.

Fix

A 5th pass after pass 4 that normalizes both sides to NFC before comparing:

return tryMatch(
  lines,
  pattern,
  startIndex,
  (a, b) => a.normalize("NFC").trim() === b.normalize("NFC").trim(),
  eof,
)

This is safe: NFC normalization followed by exact equality only matches strings Unicode defines as the same text — no false positives. (Approach proposed by the issue reporter.)

Test plan

• Added regression tests: exact match, NFD-file ↔ NFC-pattern match (Cyrillic й), and a negative case (no false match)
• packages/opencode patch suite: 23/23 pass
• No new typecheck errors in patch/index.ts (the only typecheck error on dev is a pre-existing, unrelated one in src/bus/global.ts)

Fixes #31651