fix(bigquery): support AI.FORECAST TVF parsing

[RedZapdos123]

Summary:

This PR adds BigQuery parser/generator support for AI.FORECAST table valued function syntax.

The current main parses ML.FORECAST(...) but fails on documented AI.FORECAST(...) TVF forms with named arguments.

Resolves #7456.

Changes done:

• Added a dedicated AIForecast expression:
  • sqlglot/expressions/functions.py
• Added SQL generation for AIForecast:
  • sqlglot/generator.py
  • Preserves TABLE <name> prefix for table input, while keeping parenthesized query input unchanged.
• Updated BigQuery FORECAST function parser routing:
  • sqlglot/parsers/bigquery.py
  • ML.FORECAST still uses existing ML parser path.
  • Non-ML FORECAST (for AI.FORECAST) now parses TVF first arg (TABLE <table> or (SELECT ...)) plus named args.
• Added regression coverage:
  • tests/dialects/test_bigquery.py
  • New identity checks for:
    • AI.FORECAST(TABLE citibike_trips, data_col => 'num_trips', timestamp_col => 'date', horizon => 30)
    • AI.FORECAST((SELECT * FROM citibike_trips), data_col => 'num_trips', timestamp_col => 'date', horizon => 30)

Validation:

Executed in WSL:

python -m pytest -q tests/dialects/test_bigquery.py -k "ml_functions"

Result:

• 1 passed, 57 deselected.

The screenshot of the focused test suite run, locally, on WSL for validation:

Ran the full test suite, locally on WSL for validation:

python -m pytest -q

Result:

• 1074 passed in 57.76s.

The screenshot of the full test suite run, locally, on WSL for validation:

[georgesittas]

I think this can be cleaned up a bit, but I'll take it to the finish line. Thanks for the contribution!