Lambda performance revamp

[taniabogatsch]

This PR refactors the lambda function execution code. Most notably, it removes all calls to Flatten and other 'relics' from the initial implementation. The changes significantly increase performance and make the code more structured and readable. This should make the lambdas efficient enough to be used directly in cases where unnesting is the current alternative. They can now scale to lists with a combined (over the whole table) element count of over 10M.

This also paves the way for an efficient future list_reduce implementation (@maiadegraaf, @Maxxen).

Benchmark numbers

Here are the benchmark numbers for 10K values. The benchmarks included in this PR run with 100K values. I took the benchmarks from those that @cryoEncryp used in his PR.

This PR

tania@motorbook duckdb % build/release/benchmark/benchmark_runner benchmark/micro/lambdas/list_transform_zip.benchmark   
name	run	timing
benchmark/micro/lambdas/list_transform_zip.benchmark	1	0.000477
benchmark/micro/lambdas/list_transform_zip.benchmark	2	0.000905
benchmark/micro/lambdas/list_transform_zip.benchmark	3	0.000492
benchmark/micro/lambdas/list_transform_zip.benchmark	4	0.000662
benchmark/micro/lambdas/list_transform_zip.benchmark	5	0.000542
tania@motorbook duckdb % build/release/benchmark/benchmark_runner benchmark/micro/lambdas/list_transform_select.benchmark
name	run	timing
benchmark/micro/lambdas/list_transform_select.benchmark	1	0.000461
benchmark/micro/lambdas/list_transform_select.benchmark	2	0.000336
benchmark/micro/lambdas/list_transform_select.benchmark	3	0.000233
benchmark/micro/lambdas/list_transform_select.benchmark	4	0.000287
benchmark/micro/lambdas/list_transform_select.benchmark	5	0.000279
tania@motorbook duckdb % build/release/benchmark/benchmark_runner benchmark/micro/lambdas/list_filter_where.benchmark    
name	run	timing
benchmark/micro/lambdas/list_filter_where.benchmark	1	0.000475
benchmark/micro/lambdas/list_filter_where.benchmark	2	0.000395
benchmark/micro/lambdas/list_filter_where.benchmark	3	0.000367
benchmark/micro/lambdas/list_filter_where.benchmark	4	0.000513
benchmark/micro/lambdas/list_filter_where.benchmark	5	0.000519

Feature branch

tania@motorbook duckdb % build/release/benchmark/benchmark_runner benchmark/micro/lambdas/list_transform_zip.benchmark
name	run	timing
benchmark/micro/lambdas/list_transform_zip.benchmark	1	0.852203
benchmark/micro/lambdas/list_transform_zip.benchmark	2	0.836401
benchmark/micro/lambdas/list_transform_zip.benchmark	3	0.839036
benchmark/micro/lambdas/list_transform_zip.benchmark	4	0.833708
benchmark/micro/lambdas/list_transform_zip.benchmark	5	0.843531
tania@motorbook duckdb % build/release/benchmark/benchmark_runner benchmark/micro/lambdas/list_transform_select.benchmark
name	run	timing
benchmark/micro/lambdas/list_transform_select.benchmark	1	0.444572
benchmark/micro/lambdas/list_transform_select.benchmark	2	0.438607
benchmark/micro/lambdas/list_transform_select.benchmark	3	0.445821
benchmark/micro/lambdas/list_transform_select.benchmark	4	0.431943
benchmark/micro/lambdas/list_transform_select.benchmark	5	0.439142
tania@motorbook duckdb % build/release/benchmark/benchmark_runner benchmark/micro/lambdas/list_filter_where.benchmark    
name	run	timing
benchmark/micro/lambdas/list_filter_where.benchmark	1	0.422714
benchmark/micro/lambdas/list_filter_where.benchmark	2	0.420716
benchmark/micro/lambdas/list_filter_where.benchmark	3	0.424036
benchmark/micro/lambdas/list_filter_where.benchmark	4	0.421465
benchmark/micro/lambdas/list_filter_where.benchmark	5	0.418798

SQL

# list_transform_zip
# load
CREATE TABLE tbl (l INTEGER[], zip VARCHAR[]);
INSERT INTO tbl VALUES (range(10000), [x::VARCHAR || (x + 1)::VARCHAR for x in range(10000)]);
# run
SELECT list_transform(l, (x, x_i) -> (x, zip[x_i])) FROM tbl;

# list_transform_select
#load
CREATE TABLE tbl (l INTEGER[], sel INTEGER[]);
INSERT INTO tbl VALUES (range(10000), range(10000));
# run
SELECT list_transform(sel, x -> l[x]) FROM tbl;

# list_filter_where
# load
CREATE TABLE tbl (l INTEGER[], where_l BOOLEAN[]);
INSERT INTO tbl VALUES (range(10000), [x % 13 != 0 for x in range(10000)]);
# run
SELECT list_filter(l, (x, x_i) -> where_l[x_i]) FROM tbl;

[maiadegraaf]

This PR makes me very happy :)
Not only is it more efficient, but it also is way more readable!

I just have one small formatting suggestion: could the functors be moved to their respective files so that all the filter, transform, etc., functions are together? Since we are planning on adding more lambda functions (eg. reduce), it will keep lamda_functions.cpp more concise.

[taniabogatsch]

@Mytherin, all green :) @maiadegraaf and I discussed continuing the modularization and moving the functors in a different PR, as she'll be working on list_reduce.

[Mytherin]

Thanks!