fix(arrow/array): use scale-aware ValueStr in decimal array String()

[zeroshade]

Rationale for this change

Decimal arrays stringify the raw decimal.Num struct instead of the scaled logical value. fmt.Println on a decimal array — or on a Record/RecordBatch containing one — prints e.g. {1999 0} instead of 19.99, even though ValueStr already formats the value correctly using the type's scale. This is inconsistent with ValueStr/JSON output and with other Arrow implementations (e.g. PyArrow prints 19.99).

What changes are included in this PR?

• baseDecimal[T].String() now formats each non-null value via ValueStr (which applies the type's scale through GetOneForMarshal) instead of fmt.Sprintf("%v", Value(i)). Because baseDecimal is generic, this corrects Decimal32/64/128/256 at once, and RecordBatch printing inherits the fix through the column String() path (record.go).
• Updated two existing slice-test assertions in decimal128_test.go / decimal256_test.go that pinned the old raw-struct output (they were already inconsistent with the adjacent ValueStr expectations on the next lines).
• Added TestDecimal128StringScaled, reproducing the issue (decimal128(5, 2) = 19.99) and asserting both the array String() and the RecordBatch output.

Are these changes tested?

Yes. go test ./arrow/array/ passes, including the new regression test and the updated slice tests. go build ./... and go vet/golangci-lint (via pre-commit) pass.

Are there any user-facing changes?

Yes — the textual output of String() on decimal arrays, and of any Record/RecordBatch containing decimal columns, now shows scaled decimal values (e.g. 19.99) instead of the internal struct (e.g. {1999 0}). This only affects human-readable stringification; binary/IPC/JSON representations are unchanged.

Closes #848