feat: `NaN` value in table rows

spark support Not-a-Number value (`NaN`), the same as offline storage: table data imported from parquet file.

SQL Engine do not knowns `NaN`, this lead to undefined and incorrect result for SQL queries

The issue mentions NaN support in table rows from Parquet files and SQL engine handling. Look at the Parquet data import path and SQL query execution code. Identify where NaN values are currently lost or misinterpreted. Check test files for existing NaN or floating-point handling to understand the expected behavior.

Written by the indexing model from the issue text.