Converting timestamp_ntz to array<timestamp_ntz> can cause NPE or SEGFAULT on parquet vectorized reader

Repro:

```

val path = "/tmp/sample_parquet_file"

spark.sql("SELECT CAST('2019-01-01' AS TIMESTAMP_NTZ) AS field").write.parquet(path)
spark.read.schema("field ARRAY<TIMESTAMP_NTZ>").parquet(path).collect()

```

Depending on the memory mode, it will throw an NPE on OnHeap mode and SEGFAULT on OffHeap mode.