BUGFIX: make sure symmetric mass ratio < 1 / 4
I noticed that for cases where m1 is very close to m2, bilby.gw.conversion.component_mass_to_symmetric_mass_ratio
can return eta > 0.25
by some rounding error.
This can then cause errors when converting from effective to component tidal deformabilities as they become complex.
This MR just explicitly enforces the requirement that eta <= 0.25
.