…sed columns

In SqlOptimiser.propagateTopDownColumns0(), WHERE clause and timestamp
column literals were emitted to union model branches by name resolution.
Since UNION matches columns by position, not name, the same alias (e.g.
"ts") can resolve to different column indices in different branches. This
caused one branch to receive extra top-down columns, producing a column
count mismatch that triggered the assert in
SqlCodeGenerator.checkIfSetCastIsRequired().

For example, given:
  select ts from (
    select name1, ts1, sym1 ts from t1
    union all
    select name2, ts2 ts, sym2 from t2
  ) where ts in '2025-12-01T01;2h'

The alias "ts" maps to position 2 (sym1) in the left branch but
position 1 (ts2) in the right branch. The name-based emission added the
wrong positional column to the right branch, while the subsequent
index-based propagation added the correct one, leaving the right branch
with one extra column.

The fix removes the name-based emission loops for both WHERE clause and
explicit timestamp literals. The existing index-based propagation lower
in the method already correctly handles union branches by deriving
positional indices from the left branch's top-down columns.

Co-Authored-By: Claude Opus 4.5 <[email protected]>

When a GROUP BY/SAMPLE BY model is the first UNION member and the
outer query selects a subset of columns, the optimizer adds non-
aggregate key columns to topDownColumns but doesn't propagate them
to union siblings, causing a column count mismatch assertion in
SqlCodeGenerator.checkIfSetCastIsRequired.

Propagate the same key columns to all union siblings whose
topDownColumns are already populated.

Co-Authored-By: Claude Opus 4.6 <[email protected]>