Follow PDG conventions by default for Breit–Wigner and form factors

[redeboer]

#519 and #520 add opt-in arguments for the conventions of the PDG 2026 review on resonances, but keep AmpForm's current conventions as defaults to avoid changing existing models in a patch release. This issue proposes to switch the defaults in the next minor release, so that AmpForm follows the PDG out of the box:

Expression class	Argument	Current default	New default
SimpleBreitWigner, BreitWigner	numerator	"mass-width": $m_0\Gamma_0$ in the numerator, $\lvert\hat{\mathcal{R}}^\mathrm{BW}(m_0^2)\rvert=1$	"unity": Equation (50.31)
BlattWeisskopfSquared, FormFactor	normalize	True: $\hat{B}_L^2(1)=1$	False: vertex factor $n_L$, Equations (50.33) and (50.34)

This changes the numerical output of models that rely on the defaults, and the meaning of their couplings:

• The Breit–Wigner amplitude loses its $m_0\Gamma_0$ factor, which depends on fit parameters, so it cannot be absorbed into a constant coupling.
• Form factors for $L>0$ are scaled by $1/\lvert h_L^{(1)}(1)\rvert$, that is $1/\sqrt{2}$ for $L=1$ and $1/\sqrt{13}$ for $L=2$. This affects relativistic_breit_wigner_with_ff, the channel widths of MultichannelBreitWigner, the vertex factors in the K-matrix module, and ChewMandelstamIntegral. It does not affect EnergyDependentWidth, where the constant cancels.

The builders in ampform.dynamics.builder should either adopt the new defaults or expose the arguments, and the documentation should state the convention in each case. Consider warning in the last 0.16 release when a class is constructed without an explicit argument.

Related issues

• Unify Breit-Wigner builders with explicit normalization conventions ampform-dpd#238 would then be able to rely on AmpForm's defaults for the serialized amplitude models.