Split TPCH tests in correctness, planning and explain analyze

[gabotechs]

I'm finding a bit hard to deal with the TPCH tests as they currently are:

• Sometimes, I want to just see what's the plan, and for that, I'm fine with a very low TPCH scale, and I'm fine not even executing the queries
• Sometimes, I want to ensure correctness with a very big TPCH scale, and for that I want to greatly increase the TPCH scale, and I don't want to be dealing with plans

This PR splits the TPCH tests in 3:

• plan tests: the queries get just planned, but they do not run. It's fine that there's a very small TPCH SF.
• correctness tests: no plan is compared, just the actual results between single node vs distributed.
• explain analyze tests: the plans with metrics are asserted, but it's fine if there's a small TPCH SF.