This PR seeks to resolve this issue: #2245 (comment).

• In the OrderByElement() JJTree rule, introduced a new Token collateToken to capture the literal or quoted identifier following COLLATE.
• Added a lookahead for <K_COLLATE> in the grammar so that if COLLATE appears after the expression, we wrap the existing columnReference in a new CollateExpression(columnReference, collateToken.image).
• This ensures that COLLATE 'UND-X-ICU' (or an identifier) is parsed correctly in ORDER BY clauses.

For example, the following SQL statement would fail:

select *
from a
order by cast(a.xyz as text) collate "und-x-icu" asc nulls first

with error:
Encountered unexpected token: "COLLATE" "COLLATE" at line 1, column 46. Was expecting one of: <EOF> <ST_SEMICOLON>

Now this same statement can be parsed succesfully.