feat: support trim with numeric and variant types

[seruman]

Snowflake's TRIM accepts any type and returns VARCHAR, docs does not mentioned it but it seems to only works on VARCHAR/TEXT typed inputs and acts as a cast to VARCHAR for others. DuckDB's TRIM only accepts VARCHAR. It seemed safe to me to cast input to VARCHAR to mimic similar behaviour.

select 
    trim(5) as c_number, SYSTEM$TYPEOF(c_number),
    trim(' str ') as c_varchar, SYSTEM$TYPEOF(c_varchar),
    trim(to_variant({'k': 'v'})) as c_variant, SYSTEM$TYPEOF(c_variant)
;

C_NUMBER	SYSTEM$TYPEOF(C_NUMBER)	C_VARCHAR	SYSTEM$TYPEOF(C_VARCHAR)	C_VARIANT	SYSTEM$TYPEOF(C_VARIANT)
5	VARCHAR(16777216)[LOB]	str	VARCHAR(5)[LOB]	{"k":"v"}	VARCHAR(16777216)[LOB]

[tekumara]

This fails for me on a real Snowflake instance with:

ProgrammingError: 002023 (22000): SQL compilation error:
Expression type does not match column data type, expecting VARIANT but got VARCHAR(18) for column DATA

Am guessing we need to convert it to variant first?

[seruman]

My bad again, updated.

As on #78 tests seems to fail due to -guessing- sqlalchemy version
#78 (comment)