LCFitter.fit_position fails with LCEmpiricalFourier

[fcotizelati]

LCFitter.fit_position() currently fails with an LCEmpiricalFourier template for two reasons:

1. SciPy passes a one-element array to set_overall_phase(), causing NumPy 2.x to raise:
   ValueError: setting an array element with a sequence

2. After converting the phase to a scalar, it fails because LCEmpiricalFourier does not implement the first- and second-order derivatives required by fit_position():
   NotImplementedError: No derivative function found for this object.

Suggested fix:

• explicitly convert the optimizer phase to a scalar;
• implement analytic first- and second-order derivatives for LCEmpiricalFourier;
• add a test running LCFitter.fit_position() with an empirical Fourier template.