[bugfix 1.1.1] fix bug of vhash join build

[jacktengg]

Proposed changes

Issue Number: close #10595

Problem Summary:

When build hash table in HashJoinNode::_hash_table_build, build side data is collected as a vector of Blocks of size 4G at maximum, if it happens that the total size of build side data is exactly e.g. 4G, then an empty Block will be added to the vector:

    while (!eos) {
        block.clear_column_data();
        RETURN_IF_CANCELLED(state);

        RETURN_IF_ERROR(child(1)->get_next(state, &block, &eos));
        _hash_table_mem_tracker->consume(block.allocated_bytes());
        _mem_used += block.allocated_bytes();

        if (block.rows() != 0) {
            mutable_block.merge(block);
        }

        // make one block for each 4 gigabytes
        constexpr static auto BUILD_BLOCK_MAX_SIZE = 4 * 1024UL * 1024UL * 1024UL;
        if (UNLIKELY(_mem_used - last_mem_used > BUILD_BLOCK_MAX_SIZE)) {
            _build_blocks.emplace_back(mutable_block.to_block());
            // TODO:: Rethink may we should do the proess after we recevie all build blocks ?
            // which is better.
            RETURN_IF_ERROR(_process_build_block(state, _build_blocks[index], index));

            mutable_block = MutableBlock();
            ++index;
            last_mem_used = _mem_used;
        }
    }

    _build_blocks.emplace_back(mutable_block.to_block());

Later when building RF in RuntimeFilterSlotsBase::insert, it will try to get a column from an empty Block, which causes the problem.

Also fixed a potential bug: when build side data exceed 4G * 128, the Blocks added previously into the vector may be relocated, which will invalidate the Block address used as map key in td::unordered_map<const Block*, std::vector<int>> _inserted_rows;.

Checklist(Required)

1. Does it affect the original behavior: (Yes/No/I Don't know)
2. Has unit tests been added: (Yes/No/No Need)
3. Has document been added or modified: (Yes/No/No Need)
4. Does it need to update dependencies: (Yes/No)
5. Are there any changes that cannot be rolled back: (Yes/No)

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...

[carlvinhust2012]

should use BE code formater to format the code.

[jacktengg]

done

[dataroaring]

It is better to explain how the bug happens.

[jacktengg]

Comment is added in Problem Summary.

[yiguolei]

LGTM