[fix](sql-functions) provide setup data for BITMAP_HASH / BITMAP_UNION examples

[boluor]

Both bitmap example pages query tables that the older copies never define, so a reader who runs the examples hits table does not exist.

• bitmap-hash (dev + 3.x + 2.1, EN+ZH): these copies still showed the old words example with an unreproducible expected count (33263478) and no table behind it. Port the version-4.x rewrite — a concrete words table (6 rows, 4 distinct) with the matching reproducible result (4), plus the note that a real-scale corpus returns far larger numbers. (4.x is unchanged.)
• bitmap-union (3.x + 2.1, EN+ZH): the example reads an aggregate-model pv_bitmap (the page documents that table near the bottom) but never creates it. Add a runnable -- setup that builds the aggregate table (user_id BITMAP BITMAP_UNION, AGGREGATE KEY(dt,page)) and loads the to_bitmap rows, so the dt/page result and the dedup count (3) reproduce. An aggregate model is required — a duplicate-model table would not collapse to the two rows the doc prints. (dev/4.x already carry setup.)

No rendered prose, expected output, or ja-source/ is altered beyond replacing the unreproducible bitmap-hash count with the reproducible 4.x value.

Verification — every touched example executed end-to-end on fresh single-BE clusters, reproducing the doc's printed output cell-for-cell:

version	cluster	bitmap-hash	bitmap-union
dev	master daily (doris-0.0.0-2e72603618c)	P5 F0 (EN+ZH)	unchanged (control: P4 F0)
3.x	3.1.4-rc02	P5 F0 (EN+ZH)	P3 F0 (EN+ZH)
2.1	2.1.11-rc01	P5 F0 (EN+ZH)	P3 F0 (EN+ZH)

🤖 Generated with Claude Code