[BUG] rule annotation breaks rewriting about shine HOT 3 CLOSED

1. It is not so much about the @rule, but more about the def versus val. I think in one case you create two tuning parameters, and in the other case only one.
   To check this, change the val into def topDownSplitJoin, and it will also fail to be applied twice.

2. I think that the type inference should accept 1 = 1 as being a solution without any required subsitution. Currently it is probably expected that a subsitution is required, because == was checked before running the pivot mechanism. Here, I believe that we observe that the pivot mechanism is better at deciding == than the default == of ArithExpr.

from shine.

Here is a fix: fdc0dc7

from shine.

I see, that makes sense! Didn't observe that only one tuning parameter was created. Thank you for the fix.

from shine.