Fix ExpressionVirtualColumn capabilities; fix groupBy's improper uses of StorageAdapter#getColumnCapabilities.

[gianm]

1. A usage in "isArrayAggregateApplicable" that would potentially incorrectly use
   array-based aggregation on a virtual column that shadows a real column.
2. A usage in "process" that would potentially use the more expensive multi-value
   aggregation path on a singly-valued virtual column. (No correctness issue, but
   a performance issue.)

Also makes ExpressionVirtualColumn always report that it is multi-valued. Previously,
it always set its capabilities to be singly-valued, which was bug ever since #7588, since
it might actually be multi-valued.

[jihoonson]

Thanks for fixing this. +1 after CI.

[fjy]

@gianm I think this is failing UT

[gianm]

It looks like the tests were failing because ExpressionVirtualColumn always set its capabilities to be singly-valued, which is a bug, since ever since #7588 they might be multi-valued. However, that bug was probably not detected since it was masked by this bug (which prevents groupBy from using its all-singly-valued-dimension optimization if some of the columns involved are virtual columns).

I pushed a fix for the ExpressionVirtualColumn issue and updated the top comment. In this fix I just set it to always be "true". This isn't ideal, since it means singly-valued optimizations won't work on top of it, but I didn't see an easy way for the ExpressionVirtualColumn to determine upfront if it will be singly-valued or not. I think this should be possible in the future as we add more upfront type info to the expression system, so I added a comment saying as much.

/cc @clintropolis

[clintropolis]

I pushed a fix for the ExpressionVirtualColumn issue and updated the top comment. In this fix I just set it to always be "true". This isn't ideal, since it means singly-valued optimizations won't work on top of it, but I didn't see an easy way for the ExpressionVirtualColumn to determine upfront if it will be singly-valued or not. I think this should be possible in the future as we add more upfront type info to the expression system, so I added a comment saying as much.

👍 I think this makes sense for now, I will follow this up with a fix to allow us to determine when a single input column will produce an scalar or array output so we can have this optimization again where possible.

[clintropolis]

🤘