Tracking Issue: Duckdb projection expression pushdown

[myrrc]

We want to push down scalar (and, in the future, aggregate) functions that are part of SELECT into vortex.

Example: SELECT strlen(col) doesn't need to decompress the strings.

Duckdb PR for type pushdown which is the same mechanism: duckdb/duckdb#22788

Issues to solve:

• byte_length is u64 but Duckdb wants i64. cast() decompresses the strings, although it shouldn't.
  cast reduce rule for dict evaluated validity(), causing decompression. Solved by adding validity() to byte_length() Constant comparison and byte_length OnPair kernels #8371
• Dict reader values use SharedArray which uncompresses strings into VarBinView, so we have no way to access FSST directly.
  This boils down to the task whether we want to push expression or subexpression tree to Dict values cache to avoid the cost of recanonicalization. Push down some expressions to Dict layout reader's cached values #8341

Future work:

• Push down chains of functions f_1(...(f_n(col)) if leaf child is BoundColumnRef.
• Push down multiple expressions for column if we can push all of them i.e. allow pushing down (strlen(col), col). This requires a change on projection ids passed to vortex. In this example we need to pass projection_ids=[0, 0].

Future separate extension is aggregate function pushdown

[myrrc]

This change regressed vortex-compact significantly on the same query. We can't solve this issue now as ZSTD stores lengths with data, so you need to decompress. Implementing a kernel for byte_length which only used decompressed length without constructing a VarBinView didn't yield any performance improvements. We'll accept the regression as there's no way for reader to distinguish vortex vs. vortex-compact