Inclination and phase handling

In the non-precessing case, inclination and phase parameters do not impact the frequency evolution of the signal, hence they do not need to be included during the interpolants construction. Add a flag which allows to directly skip their inclusion.

In the (l,m)=(2,2) case, this is easy to do, it suffices to fix them to a given value (incl*, phi*) and then during PE the Y_22(incl,phi) should be inserted back (exactly as done for the distance) by multiplication (after diving for Y_22(incl*, phi*)).

In the higher modes case, this can still be done, but reinserting back the spherical harmonics dependence is more cumbersome. A first option is to perform an ROQ of each (l,m) mode, which is expensive. A second option is to still fix (incl,phi) to (incl*, phi*), and then re-insert the angular dependence during PE. Since now there are multiple angular contributions, first one would need to de-convolve each mode h_{lm} = \int domega (hp - i hc) * Y_{lm} and then multiply by Y_lm each of them. This is probably costly to be done in PE.

At the moment, the employed strategy is to fix (incl, phi) when using (l,m)=(2,2) and simply include (incl,phi) in the ROQ construction when including HMs.