[fix](doc) translate.md: align en/zh to 11 examples; fix en table widths and case mixing

[boluor]

Summary

Doc page (4.x): `scalar-functions/string-functions/translate.md` (EN + ZH).

This page was Pattern-A: each side covered some corner cases the other didn't, plus EN had several small bugs.

EN bugs fixed

• Result tables for examples 1-3 had divider rows 2 chars short of the column-name row — the ASCII grid rendered ragged on docusaurus.
• Examples 1-4 used lowercase `translate(...)`, examples 5-9 used uppercase `TRANSLATE(...)` — the column headers in the result blocks were inconsistent within the same page.
• Missed four corner cases ZH already covered: multi-character mapping (`'ac' -> 'zx'`), NULL propagation, empty-string edge cases, and the duplicate-on-both-sides example (`'aabbccaa', 'abab', 'xyuv'`).

EN content trims

• Dropped the standalone "char deletion via empty ``" example — its behavior is now in the broader empty-string example.
• Dropped the Palhoça example — UTF-8 coverage is in the multi-byte example.

ZH gains

• Added the "every occurrence in source is replaced" example (`'abcda', 'a', 'z'` -> `zbcdz`) that EN had, which makes the one-occurrence-vs-many distinction explicit.
• Light prose edits so the example intros read as parallel translations.

After this change both pages share the same 11 examples in the same order. Every result block is verbatim mysql client output on Apache Doris 4.1.1.

Test plan

• Run every example end-to-end on the cluster — each result block matches.
• EN and ZH structure: identical 11-step list.
• No `Syntax` block changes.

🤖 Generated with Claude Code