feat(Duckdb): Add transpilation support for ARRAY_GENERATE_RANGE function

[fivetran-amrutabhimsenayachit]

When ARRAY_GENERATE_RANGE has a negative step value, the current transpilation produces incorrect results.
For example:

Snowflake: ARRAY_GENERATE_RANGE(-5, -25, -10) -> [-5, -15]
Current DuckDB: GENERATE_SERIES(-5, -25 - 1, -10) -> [-5, -15, -25] (WRONG)

Another example:

Snowflake: ARRAY_GENERATE_RANGE(5, 1, -1) -> [5, 4, 3, 2]
Current DuckDB: GENERATE_SERIES(5, 1 - 1, -1) -> [5, 4, 3, 2, 1, 0] (WRONG)

Fix:
Added pattern detection in the generateseries_sql function in. DuckDB:

• Detects when end argument is Sub(stop, 1) (Snowflake's pattern)
• Extracts the original stop value and uses RANGE(start, stop, step) instead
• RANGE has exclusive-end semantics matching Snowflake perfectly

DuckDb after transpilation:

"SELECT RANGE(2, 5) AS basic_range, RANGE(5, 25, 10) AS with_step, RANGE(-5, -25, -10) AS negative_step, RANGE(10, 10) AS empty_start_equals_stop, RANGE(0, 10, 2) AS even_numbers, RANGE(NULL, 5) AS null_start, RANGE(1, NULL) AS null_stop, RANGE(1, 5, NULL) AS null_step, RANGE(5, 1, -1) AS descending, RANGE(CAST(1.5 AS INT), CAST(5.8 AS INT)) AS from_float, RANGE(10, 5) AS empty_wrong_direction_positive, RANGE(-10, -5, -1) AS empty_wrong_direction_negative, RANGE(-5, 5, 2) AS negative_to_positive, RANGE(0, -10, -3) AS zero_to_negative"
┌─────────────┬───────────┬───────────────┬─────────────────────────┬─────────────────┬────────────┬───────────┬───────────┬──────────────┬──────────────┬────────────────────────────────┬────────────────────────────────┬──────────────────────┬──────────────────┐
│ basic_range │ with_step │ negative_step │ empty_start_equals_stop │  even_numbers   │ null_start │ null_stop │ null_step │  descending  │  from_float  │ empty_wrong_direction_positive │ empty_wrong_direction_negative │ negative_to_positive │ zero_to_negative │
│   int64[]   │  int64[]  │    int64[]    │         int64[]         │     int64[]     │  int64[]   │  int64[]  │  int64[]  │   int64[]    │   int64[]    │            int64[]             │            int64[]             │       int64[]        │     int64[]      │
├─────────────┼───────────┼───────────────┼─────────────────────────┼─────────────────┼────────────┼───────────┼───────────┼──────────────┼──────────────┼────────────────────────────────┼────────────────────────────────┼──────────────────────┼──────────────────┤
│ [2, 3, 4]   │ [5, 15]   │ [-5, -15]     │ []                      │ [0, 2, 4, 6, 8] │ NULL       │ NULL      │ NULL      │ [5, 4, 3, 2] │ [2, 3, 4, 5] │ []                             │ []                             │ [-5, -3, -1, 1, 3]   │ [0, -3, -6, -9]  │
└─────────────┴───────────┴───────────────┴─────────────────────────┴─────────────────┴────────────┴───────────┴───────────┴──────────────┴──────────────┴────────────────────────────────┴────────────────────────────────┴──────────────────────┴──────────────────┘

[github-actions]

SQLGlot Integration Test Results

Comparing:

• this branch (sqlglot:RD-1147618-array-generate-range, sqlglot version: RD-1147618-array-generate-range)
• baseline (main, sqlglot version: 28.10.1.dev116)

⚠️ Limited to dialects: duckdb, snowflake

By Dialect

dialect	main	sqlglot:RD-1147618-array-generate-range	transitions	links
duckdb -> duckdb	4003/4003 passed (100.0%)	4003/4003 passed (100.0%)	No change	full result / delta
snowflake -> duckdb	1579/2483 passed (63.6%)	1579/2483 passed (63.6%)	No change	full result / delta
snowflake -> snowflake	2703/2703 passed (100.0%)	2703/2703 passed (100.0%)	No change	full result / delta

Overall

main: 9189 total, 8285 passed (pass rate: 90.2%), sqlglot version: 28.10.1.dev116

sqlglot:RD-1147618-array-generate-range: 9189 total, 8285 passed (pass rate: 90.2%), sqlglot version: RD-1147618-array-generate-range

Transitions:
No change

[georgesittas]

Why are we hard-coding checks for literals? This isn't a good signal. What are you trying to check here? Should we be using an explicit arg instead?