Fix nullable type handling for column types

[sert121]

Summary

• Previously only date/datetime types were wrapped in Nullable() when the source schema allowed nulls
• Boolean, integer, float, and string columns with NULL values in the source DB were created as non-nullable in Clickhouse, silently converting NULLs to defaults (false, 0, empty string)
• Now checks the JSON schema type array for "null" and wraps any non-primary-key column in Nullable() accordingly

Root cause

The to_sql_type() method in connectors.py had explicit Nullable() wrapping for DATE, TIMESTAMP, TIME, and DATETIME types, but no handling for BOOLEAN, INTEGER, FLOAT, VARCHAR, etc. The Singer SDK JSON schema represents nullable fields as {"type": ["boolean", "null"]} but this null indicator was ignored for non-date types.

Impact

Affects any source table with nullable boolean/integer/float/string columns. Confirmed on Outdoorsy's rentals table: 18 boolean columns all created as Bool instead of Nullable(Bool), with all source NULLs converted to false.

Test plan

• Verify nullable bool column from Postgres creates Nullable(Bool) in Clickhouse
• Verify non-nullable columns remain non-nullable
• Verify primary key columns are not wrapped in Nullable