feat: add fast path for cross join to optimize tpcds q23 (~50x)

[xudong963]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

Note: tpcds 1GB. Local machine M1 mac, 8c 32GB

This PR adds a fast path for cross join when build side only has a row, which matches tpcds q23's pattern(build side is from a max/sum agg).

By the fast path, we can avoid memory reallocation. Then reduce cache miss and improve the ability of SIMD for next operators.

Before: 26s

Now: 0.4s

Tests

• [ ] Unit Test
• [ ] Logic Test
• [ ] Benchmark Test
• [x] No Test - Explain why

Type of change

• [ ] Bug Fix (non-breaking change which fixes an issue)
• [ ] New Feature (non-breaking change which adds functionality)
• [ ] Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• [ ] Documentation Update
• [ ] Refactoring
• [x] Performance Improvement
• [ ] Other (please describe):

---

This change is