Improve SQLite types handling

[AndreiKingsley]

Orginal problem in #964.

In order to support custom types not handled correctly by SQLite type affinity, custom type mapping possibility was added in #1741.

However that might not be enough. Now we only support type definitions based on sqlTypeName. However, this may not be the single determining factor.
So it seems that there are a lot of edge cases of determining the JDBC type in SQLite, and several column type determination mechanisms can be particularly relevant for it:

• column name -> type (like colTypes: Map<String, ColType> in CSV IO)
• may be column metadata (not only sqlTypeName) -> type
• inferring from values?

It's also important not only for custom types, but for simple INTEGER nullable columns .
Xerial SQLite JDBC driver seems to give identical metadata for them, however actually types of values can be both Int? and Long? (looks like a separate issue perhaps).

TableColumnMetadata(name='int_col', sqlTypeName='INTEGER', jdbcType=4, size=2147483647, javaClassName='java.lang.Object', isNullable=true)
TableColumnMetadata(name='long_col', sqlTypeName='INTEGER', jdbcType=4, size=2147483647, javaClassName='java.lang.Object', isNullable=true)

Need to investigate different usecases and and practices and improve handling of SQLite types.

[AndreiKingsley]

Added a test for nullable INTEGER column (with a small example df) in #1741