streaming processing supports nullable datatype

[lizhou1111]

This pr is for supporting nullable data type in streaming processing.Actually proton had already supported the nullable data type, but it had some bugs. Full test report:Nullable datatype in streaming processing test report.

Bug 1: group by window_start and nullable column.
For example:

drop stream if exists test;
create stream test(id nullable(int), value int);
select id, window_start, max(value) from tumble(test, 5s) group by window_start, id;
-- No matter what data you insert in any speed it won't output any result

It won’t output any result.

Bug 2: when you create external stream with nullable column, in some data type it will lose data.
For example:

drop stream if exists test_external1;
create external stream if not exists test_external1(
	int8_c int8,
	ipv4_c nullable(ipv4)
) settings type = 'kafka', brokers = 'redpanda-1:9092', topic = 'test', data_format='JSONEachRow';

select * from test_external1;
-- rpk topic produce test
-- this data is ok
{"int8_c":1,"ipv4_c":"192.168.168.168"}

-- this data will be discard
{"int8_c":2,"ipv4_c":""}

-- But if you just don't have this field, it is ok, proton will fill this field with null.
{"int8_c":3}

-- proton output:
┌─int8_c─┬─ipv4_c──────────┐
│      1 │ 192.168.168.168 │
└────────┴─────────────────┘
┌─int8_c─┬─ipv4_c─┐
│      3 │ ᴺᵁᴸᴸ   │
└────────┴────────┘

Current progress：

• Bug1 fixing
• Bug2 fixing

For bug2, if a columns are nullable, there are 4 situation about inserting data in redpanda

1. valid data({"key" : 1})
2. insert null({"key" : null})
3. insert just empty field({"key" : ""})
4. insert invalid data({"key" : "some invalid data"})

In the former proton, proton will discard the whole message for 3 and 4(bug 2), now it only happen in 4(Maybe this shouldn't happen either?).

[yl-lisen]

LGTM, don't close #422 , it's an initial PR , we can merge it first
Next, we need do more nullable type supporting or checking for functions / other processing via following PRs