[fix](agg) Avoid reusing a non-nullable column that has been converted to nullable within a block

[mrhhsg]

Proposed changes

The column that has been converted from non-nullable to nullable should not be reused.

Problem summary

0# doris::signal::(anonymous namespace)::FailureSignalHandler(int, siginfo_t*, void*) at /root/doris/be/src/common/signal_handler.h:420
 1# os::Linux::chained_handler(int, siginfo*, void*) in /usr/local/java/jdk1.8.0_202/jre/lib/amd64/server/libjvm.so
 2# JVM_handle_linux_signal in /usr/local/java/jdk1.8.0_202/jre/lib/amd64/server/libjvm.so
 3# signalHandler(int, siginfo*, void*) in /usr/local/java/jdk1.8.0_202/jre/lib/amd64/server/libjvm.so
 4# 0x00007F4051C9F400 in /lib64/libc.so.6
 5# memcpy at /root/doris/be/src/glibc-compatibility/memcpy/memcpy_x86_64.cpp:219
 6# doris::vectorized::ColumnString::deserialize_and_insert_from_arena(char const*) at /root/doris/be/src/vec/columns/column_string.cpp:226
 7# doris::vectorized::ColumnString::deserialize_vec_with_null_map(std::vector<StringRef, std::allocator<StringRef> >&, unsigned long, unsigned char const*) at /root/doris/be/src/vec/columns/column_string.cpp:283
 8# void doris::vectorized::AggregationNode::_serialize_with_serialized_key_result(doris::RuntimeState*, doris::vectorized::Block*, bool*)::{lambda(auto:1&&)#1}::operator()<doris::vectorized::AggregationMethodSerialized<PHHashMap<StringRef, char*, DefaultHash<StringRef, void>, false> >&>(doris::vectorized::AggregationMethodSerialized<PHHashMap<StringRef, char*, DefaultHash<StringRef, void>, false> >&) const at /root/doris/be/src/vec/exec/vaggregation_node.cpp:1232
 9# doris::vectorized::AggregationNode::_serialize_with_serialized_key_result(doris::RuntimeState*, doris::vectorized::Block*, bool*) at /root/doris/be/src/vec/exec/vaggregation_node.cpp:1294
10# std::_Function_handler<doris::Status (doris::RuntimeState*, doris::vectorized::Block*, bool*), std::_Bind_result<doris::Status, doris::Status (doris::vectorized::AggregationNode::*(doris::vectorized::AggregationNode*, std::_Placeholder<1>, std::_Placeholder<2>, std::_Placeholder<3>))(doris::RuntimeState*, doris::vectorized::Block*, bool*)> >::_M_invoke(std::_Any_data const&, doris::RuntimeState*&&, doris::vectorized::Block*&&, bool*&&) at /var/local/ldb-toolchain/include/c++/11/bits/std_function.h:293
11# doris::vectorized::AggregationNode::get_next(doris::RuntimeState*, doris::vectorized::Block*, bool*) at /root/doris/be/src/vec/exec/vaggregation_node.cpp:508
12# doris::ExecNode::get_next_after_projects(doris::RuntimeState*, doris::vectorized::Block*, bool*) at /root/doris/be/src/exec/exec_node.cpp:852
13# doris::PlanFragmentExecutor::get_vectorized_internal(doris::vectorized::Block**) at /root/doris/be/src/runtime/plan_fragment_executor.cpp:352
14# doris::PlanFragmentExecutor::open_vectorized_internal() at /root/doris/be/src/runtime/plan_fragment_executor.cpp:300
15# doris::PlanFragmentExecutor::open() at /root/doris/be/src/runtime/plan_fragment_executor.cpp:253
16# doris::FragmentExecState::execute() at /root/doris/be/src/runtime/fragment_mgr.cpp:251
17# doris::FragmentMgr::_exec_actual(std::shared_ptr<doris::FragmentExecState>, std::function<void (doris::PlanFragmentExecutor*)>) at /root/doris/be/src/runtime/fragment_mgr.cpp:498
18# std::_Function_handler<void (), doris::FragmentMgr::exec_plan_fragment(doris::TExecPlanFragmentParams const&, std::function<void (doris::PlanFragmentExecutor*)>)::{lambda()#1}>::_M_invoke(std::_Any_data const&) at /var/local/ldb-toolchain/include/c++/11/bits/std_function.h:291
19# doris::ThreadPool::dispatch_thread() at /root/doris/be/src/util/threadpool.cpp:542
20# doris::Thread::supervise_thread(void*) at /root/doris/be/src/util/thread.cpp:455
21# start_thread in /lib64/libpthread.so.0
22# clone in /lib64/libc.so.6

Checklist(Required)

• Does it affect the original behavior
• Has unit tests been added
• Has document been added or modified
• Does it need to update dependencies
• Is this PR support rollback (If NO, please explain WHY)

Further comments

If this is a relatively large or complex change, kick off the discussion at dev@doris.apache.org by explaining why you chose the solution you did and what alternatives you considered, etc...

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[mrhhsg]

run buildall

[hello-stephen]

TeamCity pipeline, clickbench performance test result:
the sum of best hot time: 35.94 seconds
stream load tsv: 465 seconds loaded 74807831229 Bytes, about 153 MB/s
stream load json: 23 seconds loaded 2358488459 Bytes, about 97 MB/s
stream load orc: 75 seconds loaded 1101869774 Bytes, about 14 MB/s
stream load parquet: 31 seconds loaded 861443392 Bytes, about 26 MB/s
https://doris-community-test-1308700295.cos.ap-hongkong.myqcloud.com/tmp/20230320090350_clickbench_pr_117255.html

[yiguolei]

LGTM