ALTER to LowCardinality was leading to segfault on empty parts

[filimonov]

I hereby agree to the terms of the CLA available at: https://yandex.ru/legal/cla/?lang=en

Changelog category (leave one):

• Non-significant (changelog entry is not needed)

Detailed description (optional):
When empty part was altered to LowCardinality segmentation fault was happening with a stack trace like that:

2019.11.15 12:11:14.565554 [ 392 ] {} <Error> BaseDaemon: ########################################
2019.11.15 12:11:14.565607 [ 392 ] {} <Error> BaseDaemon: (version 19.11.12.69 (official build)) (from thread 41) Received signal Segmentation fault (11).
2019.11.15 12:11:14.565639 [ 392 ] {} <Error> BaseDaemon: Address: NULL pointer. Access: read. Address not mapped to object.
2019.11.15 12:11:14.606264 [ 392 ] {} <Error> BaseDaemon: 0. /usr/bin/clickhouse-server(StackTrace::StackTrace(ucontext_t const&)+0x31) [0x7dcc1f1]
1. /usr/bin/clickhouse-server() [0x3d15f3e]
2. /lib64/libpthread.so.0(+0xf5d0) [0x7f2109a805d0]
3. /usr/bin/clickhouse-server() [0x6bb8bf3]
4. /usr/bin/clickhouse-server(DB::DataTypeLowCardinality::serializeBinaryBulkStateSuffix(DB::IDataType::SerializeBinaryBulkSettings&, std::shared_ptr<DB::IDataType::Seria
5. /usr/bin/clickhouse-server(DB::MergedColumnOnlyOutputStream::writeSuffixAndGetChecksums()+0x1f6) [0x7010966]
6. /usr/bin/clickhouse-server(DB::MergeTreeData::alterDataPart(DB::NamesAndTypesList const&, std::vector<std::shared_ptr<DB::ASTIndexDeclaration>, std::allocator<std::sha
7. /usr/bin/clickhouse-server(ThreadPoolImpl<ThreadFromGlobalPool>::worker(std::_List_iterator<ThreadFromGlobalPool>)+0x1a7) [0x772b977]
8. /usr/bin/clickhouse-server(ThreadFromGlobalPool::ThreadFromGlobalPool<ThreadPoolImpl<ThreadFromGlobalPool>::scheduleImpl<void>(std::function<void ()>, int, std::option
9. /usr/bin/clickhouse-server(ThreadPoolImpl<std::thread>::worker(std::_List_iterator<std::thread>)+0x1a6) [0x7729486]
10. /usr/bin/clickhouse-server() [0xb89e440]
11. /lib64/libpthread.so.0(+0x7dd5) [0x7f2109a78dd5]
12. /lib64/libc.so.6(clone+0x6d) [0x7f210929c02d

It like the root of the problem was fixed between 19.11.12 and 19.11.13.
Most probably it was that fix: #7062
It was fixed in #6780

Adding test just in case.

It looks like those 2 issues can be also closed: #5289 and #6877

[alexey-milovidov]

This will make the test flacky.

[alexey-milovidov]

Look at the examples for waiting for mutation to finish in another tests.

[alexey-milovidov]

Fixed.