Add quantile(), params() for parametric aggregates

[theory]

ClickHouse provides a number of "parametric aggregates" which take one or more parameters in a first set of parentheses and then normal arguments in a second. An example is quantile, which can be called like so:

SELECT quantile(0.25)(val) FROM t;

To mimic this feature, implement a params data type and function that takes a list of values of any type. Detect when it's the first argument to any aggregates and pass it as ClickHouse parameters before any other arguments. The equivalent of the above query in PostgreSQL is:

SELECT quantile(params(0.25), val) FROM t;

The pattern is general and can be used for any aggregate functions, but used here only for the new quantile and quantileExact aggregate functions. Others can be added in the future; all they have to do is use the clickhouse_func_push_fail function for their SFUNC, with one variant that takes params as the first argument and the remaining arguments specified as usual (likely as "any" to allow data type enforcement to be pushed down to ClickHouse).

Done by using the new ch_get_params_function() function to detect params()as the first argument to an aggregate function indeparseAggref()` and emitting it before appending the arguments opening parenthesis.

Document the new functions and explain how to use them.

While at it, simplify the handling of F_REGEXP_LIKE_TEXT_TEXT in Postgres 14 and earlier by simply setting it on those versions, rather than using #define to omit any code that uses it. This aligns with how the type mapper handles other such data type constants.

Also:

• Fix an Assert for a variadic array to apply to the array, rather than the parent node. * Also, remove the unused func_arg field from the deparse_expr_cxt struct and add a comment explaining the array_as_tuple field.

[theory]

Requesting another review because I came up with a better, more general approach to supporting parameterized aggregates.

[serprex]

it's hard since it's always going to be awkward how to map (args1)(args2) to postgres. this way at least makes it clearer there's something magical happening

[theory]

Yes, exactly.