Add statistics collection for added column

[pajaks]

Description

Currently when column is added when statistics are present total size is not collected for newly added column.
NDV statistics works fine.

Currently:

create table test (varchar_1 VARCHAR);
insert into test values ('a');
analyze test;
alter table test add column varchar_2 varchar;
insert into test values ('b','c');
analyze test;
show stats for test;

 column_name | data_size | distinct_values_count | nulls_fraction | row_count | low_value | high_value
-------------+-----------+-----------------------+----------------+-----------+-----------+------------
 varchar_1   |       2.0 |                   2.0 |            0.0 |      NULL | NULL      | NULL
 varchar_2   |      NULL |                   1.0 |            0.1 |      NULL | NULL      | NULL
 NULL        |      NULL |                  NULL |           NULL |       2.0 | NULL      | NULL

After change:

 column_name | data_size | distinct_values_count | nulls_fraction | row_count | low_value | high_value
-------------+-----------+-----------------------+----------------+-----------+-----------+------------
 varchar_1   |       2.0 |                   2.0 |            0.0 |      NULL | NULL      | NULL
 varchar_2   |      1.0  |                   1.0 |            0.1 |      NULL | NULL      | NULL
 NULL        |      NULL |                  NULL |           NULL |       2.0 | NULL      | NULL

Additional context and related issues

Release notes

( ) This is not user-visible or docs only and no release notes are required.
( ) Release notes are required, please propose a release note for me.
( ) Release notes are required, with the following suggested text:

# Delta Lake
* Collect statistics for newly added column

[pajaks]

Check for previous statistics of total size is probably redundant. If column was not selected in ANALYZE WITH query it should be already filtered by .filter(columnMetadata -> analyzeColumnNames.contains(columnMetadata.getName()))
I could not find any other case when those statistics are not present except for case when new column is added.

[alexjo2144]

The other situation would be if you analyzed a table before we collected TOTAL_SIZE_IN_BYTES, then upgraded Trino, and then analyzed the table again/did an insert.

That was added in 387 (June 2022), @ebyhr @findepi think that's long enough?

[findepi]

good point, Alex. let's assume this is fine.

[alexjo2144]

What do you think about instead modifying the addColumn code to add the new column to the extended stats file with a data size of zero?

[pajaks]

What do you think about instead modifying the addColumn code to add the new column to the extended stats file with a data size of zero?

That's an option. But if there is no scenario when totalSizeStatisticsExists is useful removing it will leave code less complex.

[pajaks]

What do you think about instead modifying the addColumn code to add the new column to the extended stats file with a data size of zero?

That's an option. But if there is no scenario when totalSizeStatisticsExists is useful removing it will leave code less complex.

It will also start showing 0 instead of null in case of using ANALYZE x WITH (columns = ...). I'm not sure it's a big issue but it can be misleading.

[findepi]

@alexjo2144 PTAL

[alexjo2144]

If Yuya and Piotr are okay with this: #16109 (comment) then this looks good to me.

[pajaks]

Rebased to include fix for skipped tests in CI