Bloom filters for i8 and i16 always return false negatives

**Describe the bug**

It is unclear to me if this is an issue when building or checking the Bloom filter; but either way, building a Bloom filter with `i8` or `i16` values (as opposed to `i32` or `i64`) always returns `false` when checked.

**To Reproduce**

```diff
diff --git a/parquet/src/arrow/arrow_writer/mod.rs b/parquet/src/arrow/arrow_writer/mod.rs
index 18c8617e07..d6b14e2899 100644
--- a/parquet/src/arrow/arrow_writer/mod.rs
+++ b/parquet/src/arrow/arrow_writer/mod.rs
@@ -2039,6 +2039,36 @@ mod tests {
values_required::(many_vecs_iter);
}

+ #[test]
+ fn i8_column_bloom_filter() {
+ let array = Arc::new(Int8Array::from_iter(0..SMALL_SIZE as i8));
+ let mut options = RoundTripOptions::new(array, false);
+ options.bloom_filter = true;
+
+ let files = one_column_roundtrip_with_options(options);
+ check_bloom_filter(
+ files,
+ "col".to_string(),
+ (0..SMALL_SIZE as i8).collect(),
+ (SMALL_SIZE as i8 + 1..SMALL_SIZE as i8 + 10).collect(),
+ );
+ }
+
+ #[test]
+ fn i16_column_bloom_filter() {
+ let array = Arc::new(Int16Array::from_iter(0..SMALL_SIZE as i16));
+ let mut options = RoundTripOptions::new(array, false);
+ options.bloom_filter = true;
+
+ let files = one_column_roundtrip_with_options(options);
+ check_bloom_filter(
+ files,
+ "col".to_string(),
+ (0..SMALL_SIZE as i16).collect(),
+ (SMALL_SIZE as i16 + 1..SMALL_SIZE as i16 + 10).collect(),
+ );
+ }
+
#[test]
fn i32_column_bloom_filter() {
let array = Arc::new(Int32Array::from_iter(0..SMALL_SIZE as i32));
@@ -2054,6 +2084,21 @@ mod tests {
);
}

+ #[test]
+ fn i64_column_bloom_filter() {
+ let array = Arc::new(Int64Array::from_iter(0..SMALL_SIZE as i64));
+ let mut options = RoundTripOptions::new(array, false);
+ options.bloom_filter = true;
+
+ let files = one_column_roundtrip_with_options(options);
+ check_bloom_filter(
+ files,
+ "col".to_string(),
+ (0..SMALL_SIZE as i64).collect(),
+ (SMALL_SIZE as i64 + 1..SMALL_SIZE as i64 + 10).collect(),
+ );
+ }
+
#[test]
fn binary_column_bloom_filter() {
let one_vec: Vec = (0..SMALL_SIZE as u8).collect();
```

returns:

```
failures:

---- arrow::arrow_writer::tests::i16_column_bloom_filter stdout ----
thread 'arrow::arrow_writer::tests::i16_column_bloom_filter' panicked at parquet/src/arrow/arrow_writer/mod.rs:1792:17:
Value [0, 0] should be in bloom filter

---- arrow::arrow_writer::tests::i8_column_bloom_filter stdout ----
thread 'arrow::arrow_writer::tests::i8_column_bloom_filter' panicked at parquet/src/arrow/arrow_writer/mod.rs:1792:17:
Value [0] should be in bloom filter

failures:
arrow::arrow_writer::tests::i16_column_bloom_filter
arrow::arrow_writer::tests::i8_column_bloom_filter
```

**Expected behavior**
These tests should pass

**Additional context**

I found this from Datafusion: https://github.com/apache/arrow-datafusion/issues/9779