Add the aggregate functions `BIT_AND` / `BIT_OR` / `BOOL_AND`

### What & why

The aggregate folder `udf/table/agg/` has 20 functions (`Count`, `Sum*`, `Avg*`,
`Percentile*`, …) but no bitwise aggregates (`BIT_AND` / `BIT_OR`) or boolean
aggregate (`BOOL_AND` — true only if every value is true). Aggregates are a step
up from scalar functions: you implement an `Accumulator` and, crucially, a
correct distributed `merge`. This is the issue that teaches you how distributed
aggregation actually works.

聚合函数目录已有 20 个,但缺位运算聚合 `BIT_AND`/`BIT_OR` 和布尔聚合 `BOOL_AND`
(全为真才为真)。聚合函数比标量函数稍难——要实现 `Accumulator` 并写对分布式 `merge`。
做完这题,你就真正理解了分布式聚合是怎么回事。

### The task

- `bit_and(x)` / `bit_or(x)` → bitwise AND/OR over a column of integers
- `bool_and(b)` → whether a column of booleans is all true

### Where to look

1. `udf/table/agg/Count.java` is the simplest UDAF template (it has the inner
`Accumulator` class). Implement the five methods:
`createAccumulator / accumulate / merge / resetAccumulator / getValue`.
2. The `Accumulator` **must** `implement Serializable` (it gets serialized across the cluster).
3. Register in `BuildInSqlFunctionTable.java`.
4. Build multi-row test data; ideally exercise a multi-partition `merge`.

### Done when

- [ ] Three UDAFs implemented, with **correct `merge` logic** (this is what makes distributed results correct)
- [ ] Accumulator is serializable
- [ ] Empty-input / all-null return values are defined
- [ ] Tests pass, including multi-row aggregation