[WIP] Add ORC support for the Iceberg Connector

[lxynov]

Although the spec for ORC with Iceberg has not been finalized (https://iceberg.apache.org/spec/#orc is expected to change), the comments under apache/iceberg#227 show the direction. The problem to solve is that Iceberg tracks columns by IDs and ORC tracks them by names, which makes ORC not directly workable in Iceberg's semantics. According to the discussion in the Iceberg community, the current idea is to introduce user-defined type annotations to ORC (apache/orc#410), and use them to store Iceberg column IDs.

This WIP PR prototypes this idea in Presto.

Tests done:

1. TestIcebergSmoke passes with the ORC format.
2. End-to-end test on schema evolution.

CREATE TABLE iceberg.u_xinlin.exp (order_id int, price, date varchar) WITH (format = 'orc', partitioning = array['date']);
INSERT INTO iceberg.u_xinlin.exp VALUES (0, 1.0, '2019-08-13');
SELECT * FROM iceberg.u_xinlin.exp;

It shows

 order_id | price |    date    
----------+-------+------------
        0 |   1.0 | 2019-08-13 
(1 row)

ALTER TABLE iceberg.u_xinlin.exp DROP COLUMN price;
ALTER TABLE iceberg.u_xinlin.exp ADD COLUMN price double;
SELECT * FROM iceberg.u_xinlin.exp;

It shows

 order_id |    date    | price 
----------+------------+-------
        0 | 2019-08-13 |  NULL 

This is consistent with Iceberg's schema evolution semantics.

cc: @wagnermarkd @electrum

[ebyhr]

@lxynov I confirmed you sent new PR to support ORC on Iceberg. Can we close this PR?

[lxynov]

@ebyhr Yes