FEAT - Add fit_intercept support to LBFGS solver

[floriankozikowski]

Description of the feature

The LBFGS solver currently doesn't support fit_intercept=True. A use case of this was described in Issue #320

Considered alternatives

Manual data centering works but complicates the optimization and user experience. In many use cases sklearn is currently is stil the faster alternative.

Additional context

np.random.seed(42)
X = np.random.randn(1000, 50)
y = np.random.poisson(np.exp(X.sum(axis=1) * 0.01 - 1))

# This should work but currently fails as its not supported
estimator = GeneralizedLinearEstimator(
    datafit=Poisson(),
    penalty=L2(alpha=0.01),
    solver=LBFGS(fit_intercept=True)  # <- Currently not supported
)
estimator.fit(X, y)

# Compare with sklearn (works fine)
sklearn_model = PoissonRegressor(alpha=0.01, fit_intercept=True, solver='lbfgs')
sklearn_model.fit(X, y)

Note: Pattern exists in anderson_cd.py (lines 163-166) and other solvers to see how to implement fit_intercept method.