change the data processing order in changelog

[lizhou1111]

change the data processing order in changelog.
For example:

drop stream if exists test_kv;
create stream if not exists test_kv(id int, value int) primary key id settings mode='changelog_kv';
select count_distinct(value) from test_kv;
insert into test_kv(id, value, _tp_delta) values(2, 1, 1)(2, 1, -1)(3, 2, 1)(3, 2, -1);

The output will be 2, it should be 0.Because a piece of data is inserted and then deleted, there should be no data.The sum_distinct will encountered the same problem

The previous processing strategy was to divide two chunks according to the delta value. The data with a delta value of -1 is placed in the first chunk, and the data with a delta value of 1 is placed in the second chunk(take the example above, it will be [(2, 1, -1), (3, 2, -1),(2, 1, 1), (3, 2, 1)]). Therefore, the data with a delta value of -1 will be placed in the first chunk. will be processed first, which is inconsistent with the order of data input.

The current strategy is putting consecutive data with the same _tp_delta value in a chunk.
If the input chunk delta flags are [1, 1, 1, -1, -1, 1, 1, 1].We will split into 3 chunks:[[1, 1, 1], [-1, -1], [1, 1, 1]].
This not only ensures that the order of data processing is consistent with the input, but also ensures that the _tp_delta values in the same chunk are the same.

this closes #448