feat(snowflake)!: Transpilation support for Snowflake REGEXP_COUNT to DuckDB

[fivetran-kwoodbeck]

Adds transpilation for Snowflake's REGEXP_COUNT function to DuckDB. It uses DuckDB's REGEXP_EXTRACT_ALL, which requires flags embedded in pattern as (?ims), not as a separate argument.

[github-actions]

SQLGlot Integration Test Results

Comparing:

• this branch (sqlglot:feature/transpile-regex-count, sqlglot version: feature/transpile-regex-count)
• baseline (main, sqlglot version: 28.10.1.dev82)

⚠️ Limited to dialects: duckdb, snowflake

By Dialect

dialect	main	sqlglot:feature/transpile-regex-count	transitions	links
duckdb -> duckdb	4003/4003 passed (100.0%)	4003/4003 passed (100.0%)	No change	full result / delta
snowflake -> duckdb	1553/2449 passed (63.4%)	1560/2449 passed (63.7%)	7 fail -> pass	full result / delta
snowflake -> snowflake	2661/2669 passed (99.7%)	2661/2669 passed (99.7%)	No change	full result / delta

Overall

main: 9121 total, 8217 passed (pass rate: 90.1%), sqlglot version: 28.10.1.dev82

sqlglot:feature/transpile-regex-count: 9121 total, 8224 passed (pass rate: 90.2%), sqlglot version: feature/transpile-regex-count

Transitions:
7 fail -> pass

[georgesittas]

Let's add tests.

[georgesittas]

Should this branch be consolidated with the one in L3220? So, we'd pass the supported flags wherever this helper is used.

[fivetran-kwoodbeck]

will do, yes, it's better to explicitly force the flags that are supported.

[georgesittas]

What happens if the pattern already contains the flags in validated_flags? Does it cause any issues?

[fivetran-kwoodbeck]

I don't think that would be a valid query in Snowflake, something like REGEXP_COUNT(text, '(?i)hello', 1, 'i') won't even run. The flags for REGEXP_COUNT are meant to be in parameters.

Also, on DuckDB side, prepending another one seems to run:

select REGEXP_EXTRACT_ALL('Hello World', '(?im)L');
select REGEXP_EXTRACT_ALL('Hello World', '(?i)(?im)L');