test: remove experimental multivariate test (verified univariate-only)

Experimental test confirmed TimesFM is univariate-only (raises "TimesFM only
supports univariate series"). This is now properly documented in user guide.

Removed test to keep test suite clean - the univariate validation is already
tested through the implementation.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Author: nehalecky

darts/tests/models/forecasting/test_timesfm_model.py

@@ -354,42 +354,3 @@ def test_predict_time_index_continuation(self):
         # Check that forecast time index starts after series ends
         assert forecast.start_time() == series.end_time() + series.freq
         assert len(forecast) == 10
-
-
-class TestTimesFMModelMultivariateExperimental:
-    """Experimental tests to determine TimesFM's actual multivariate capabilities"""
-
-    def test_multivariate_series_direct_forecasting(self):
-        """EXPERIMENTAL: Test if TimesFM can forecast multivariate series directly"""
-        from darts.models.forecasting.foundation import TimesFMModel
-        import numpy as np
-
-        # Create a true multivariate series (2 components)
-        multivariate_series = tg.sine_timeseries(length=100, value_frequency=0.1, column_name="A")
-        multivariate_series = multivariate_series.stack(
-            tg.sine_timeseries(length=100, value_frequency=0.2, column_name="B")
-        )
-
-        model = TimesFMModel(zero_shot=True, max_context_length=512)
-
-        # Try to forecast the multivariate series directly
-        try:
-            forecast = model.predict(n=12, series=multivariate_series)
-
-            # If this succeeds, TimesFM supports multivariate!
-            assert isinstance(forecast, TimeSeries)
-            assert len(forecast) == 12
-            assert forecast.n_components == 2  # Should preserve multivariate structure
-            assert not np.any(np.isnan(forecast.values()))
-
-            print("✅ TimesFM DOES support multivariate forecasting!")
-            print(f"   Input: {multivariate_series.n_components} components")
-            print(f"   Output: {forecast.n_components} components")
-
-        except (ValueError, TypeError, RuntimeError) as e:
-            # If this fails, TimesFM is univariate-only
-            print("❌ TimesFM does NOT support multivariate forecasting")
-            print(f"   Error: {str(e)}")
-            # This test documents the limitation - we expect it to fail
-            import pytest
-            pytest.skip(f"TimesFM is univariate-only: {str(e)}")