Huh?! I've never seen that trig ID before (because it is incorrect). I've seen [itex]cos^2(x)=\frac{1}{2}(1-cos(2x)[/itex] before though. There is an easier way though; the integral of cos^2 over a full period represents the area under cos^2 from 0 to T. Due to symmetry, the area under sin^2 will be the same so
[tex]\int_0^T cos^2(\omega(t-r/c))dt=\frac{1}{2} \left(\int_0^T cos^2(\omega(t-r/c))dt +\int_0^T sin^2(\omega(t-r/c))dt \right)[/tex]
[tex]=\frac{1}{2} \int_0^T (cos^2(\omega(t-r/c)) + sin^2(\omega(t-r/c)))dt=\frac{1}{2} \int_0^T (1)dt=\frac{T}{2}[/tex]
[tex]\Rightarrow \left< cos^2(\omega(t-r/c)) \right> _T=\frac{1}{2}[/tex]
I think you are probably supposed to take a spatial average of P_rad as well, in that case, just integrate the r,theta and phi dependent terms over the volume of a sphere and divide by the volume.