[fix](doc) correct AGG_STATE example and workload-group sql fence rendering

[boluor]

Summary

Two long-open doc bugs fixed in one PR.

1. AGG_STATE example uses combinators on a non-agg_state column (#2929)

table-design/data-model/aggregate.md declares the example table with v1 int SUM, v2 agg_state<group_concat(string)>. The page then incorrectly calls sum_union(v1) (union-insert) and sum_merge(v1) (result query) — but v1 is a plain aggregate-key column, not agg_state<sum(int)>. MergeCombinator and UnionCombinator in fe/fe-core/src/main/java/org/apache/doris/nereids/trees/expressions/functions/combinator/ cast their argument to AggStateType, so the example as written fails at analysis time.

Replace with plain sum(v1) across the union-insert and the two result queries. Numeric output (20 / 16) is unchanged and matches what the doc already claims. group_concat_merge(v2) / group_concat_union(v2) on v2 remain correct because v2 is genuinely declared as agg_state.

Result tables re-flowed to mysql-CLI-correct column widths.

2. workload-group "View Schema" SQL block renders as literal text (#2497)

admin-manual/workload-management/workload-group.md had a 4-space-indented \``sql DESC FUNCTION s3 (...) ```` block sitting directly after a plain paragraph (no parent list item). CommonMark treats 4-space indent as an indented code block, so the inner fence markers render as literal text on the rendered page — exactly what the reporter's screenshot shows.

De-indent the block to column 0. current and version-4.x already carry this fix; backport to version-2.1 and version-3.x EN + zh.

Scope

Fix	Files
AGG_STATE — sum_merge/sum_union(v1) → sum(v1)	aggregate.md × 8 (current/2.1/3.x/4.x EN+zh)
workload-group — de-indent DESC FUNCTION block	workload-group.md × 4 (2.1/3.x EN+zh)

Test plan

• Dead-link check passes
• Result tables align in monospace
• No remaining sum_merge(v1) / sum_union(v1) in aggregate.md

Closes #2497
Closes #2929