Support alternate format for Utf8 unparsing (CHAR)

[sgrebnov]

Which issue does this PR close?

PR addresses Utf8 unparser issue producing invalid CAST(col AS TEXT) SQL for MySQL.

Rationale for this change

MySQL cast function does not support TEXT and requires CHAR for CAST (automatically returns VARCHAR, TEXT, LONGTEXT)

CHAR[(N)] [charset_info]
Produces a string with the VARCHAR data type, unless the expression expr is empty (zero length), in which case the result type is CHAR(0). If the optional length N is given, CHAR(N) causes the cast to use no more than N characters of the argument. No padding occurs for values shorter than N characters. If the optional length N is not given, MySQL calculates the maximum length from the expression. If the supplied or calculated length is greater than an internal threshold, the result type is TEXT. If the length is still too long, the result type is LONGTEXT.

Example query:

 select * from customer where c_custkey = 'building

Before this change (fails in MySQL)

SELECT `customer`.`c_custkey`, ... FROM `customer` WHERE (CAST(`customer`.`c_custkey` AS TEXT) = 'building')

After this change (works in MySQL)

SELECT `customer`.`c_custkey`,  ... FROM `customer` WHERE (CAST(`customer`.`c_custkey` AS CHAR) = 'building')

What changes are included in this PR?

PR introduces configurable use_char_for_utf8_cast dialect parameter that controls whether CHAR vs TEXT/VARCHAR is used for Utf8 unparsing.

Are these changes tested?

Yes, added unit tests + manual testing

Are there any user-facing changes?

CustomDialectBuilder now supports use_char_for_utf8_cast that can be used to specify whether CHAR vs TEXT/VARCHAR data type should be used for Utf8 unparsing.

[alamb]

Thanks for the contribution @sgrebnov . I had a suggestion on the API -- let me know what you think

[sgrebnov]

I had a suggestion on the API -- let me know what you think

@alamb - 👍 I was thinking about this as well, but was unable to come up with other good potential types for string unparsing (asf data types) so decided to proceed with this approach. As we have more people thinking that we should better to return the type here I'll update my PR.

Note - as we have different types for Utf8 and LargeUtf8 there will be two functions/settings.

[alamb]

Looks good to me -- thank you @sgrebnov