[YCQL] Support selecting sub-field of a user-defined type

[m-iancu]

Jira Link: DB-4680
For example, given the following schema:

create type udt1(i int, t text);
create type udt2(i int, u1 udt1);
create table test(k int primary key, u1 udt1, u2 udt2);

Allow queries like:

SELECT u1.i from test;
SELECT u2.u1.t, u1.t from test;

These can be handled similarly to the subscript_args (used for e.g. m['x'] or l[2]['x']) and may even be able to reuse most of that code (except for the parsing stage).

As suggested by @nocaway, this should also add more tests for expressions involving (nested) udt's in general. E.g.:

• SELECT tojson(nesting_udt) FROM ....
• SELECT indexing_expression(nesting_udt) FROM ....

For example:

cqlsh:test> create type udt1(i int, t text);
cqlsh:test> create type udt2(i int, u1 frozen<udt1>);
cqlsh:test> create table tab2(k int primary key, su set<frozen<udt2>>);
cqlsh:test> insert into tab2 (k, su) values (111, { { i : 1, u1 : { i : 1, t : '1001' } }, { i : 2, u1 : { i : 2, t : '1002' } } });
cqlsh:test> insert into tab2 (k, su) values (777, { { i : 3, u1 : { i : 3, t : '1003' } }, { i : 4, u1 : { i : 4, t : '1004' } } });
cqlsh:test> select tojson(su) from tab2;

 server_tojson(su)
---------------------------------------------------
 [{"i":3,"u1":[3,"1003"]},{"i":4,"u1":[4,"1004"]}]
 [{"i":1,"u1":[1,"1001"]},{"i":2,"u1":[2,"1002"]}]

(2 rows)