Fixes #5614

This was a tough one.

When the where binder attempted to bind the 'text' alias in the where clause, the column_alias_binder would bind it with a depth of 0 (something I added a long time ago). This was incorrect. We should be binding c0 with the correct depth count.

When it has a depth of 0, it is not planned as a correlated subquery. With a depth of 1 it is, this will fix the issue, but it would break some other correlated subquery tests. Specifically tests with a having clause.

Thankfully, the tests were only broken in such a way that the error was incorrect. After looking at the code, I decided to propagate up correlated subquery binding errors instead of stalling them. The previous flow was the following

• attempt to bind an expression as if it wasn't a correlated subquery,
• if some error X is encountered, attempt to bind the expression as if it is in a correlated subquery,
• if some error Y is encountered, continue to attempt correlatedSubquery binding until there are no more context Binders.
• If the expression cannot be bound, throw error X.

The new flow will return error Y (or the last seen error from correlatedSubquery binding).

This has the benefit of fixing some lateral join errors that were a bit cryptic