I came across the following inequality: $$ \cos x + \frac{2p}{\pi}\sin x \geq 1 - 2 \left(\frac{x}{\pi}\right)^p, $$ which should be valid for any $0\leq x \leq \pi$ and $p\in\mathbb R$ according to the graphs.
I managed to prove it myself for $p\leq0$ as follows:
If $p=0$ it's trivial, and if $p<0$, I look at the difference $$f(x) = \cos x + \frac{2p}{\pi}\sin x - 1 + 2\left(\frac{x}{\pi}\right)^p$$ Noticing that $f(\pi)=0$, it is enough to show $-f'(x)\geq0$. Then, $$-f'(x) = \sin x - \frac{2p}{\pi}\left(\cos x + \left(\frac{x}{\pi}\right)^{p - 1}\right)$$ and all terms are nonnegative for $x\in[0,\frac{\pi}{2}]$. If $x\in[\frac{\pi}{2},\pi]$ then still $\cos x + \left(\frac{x}{\pi}\right)^{p - 1}\geq0$ as the latter term is at least $1$.
This method doesn't help me for the case $p>0$, though, as the function is no longer monotone. I tried to calculate (among other things) a taylor and fourier series, but found it wasn't much use for general $p$.
Any help or hints will be appreciated.
This is a partial solution or rather a long comment to give some ideas. It is probably better to compare the function $\phi_p(x)=\cos x+\frac{2p}{\pi}\sin x$ and $f_p(x)=1-2(x/\pi)^p$.
$\phi'_p(x)=-\sin x+\frac{2p}{\pi}\cos x=0$ on $[0,\pi]$ iff $\tan(x)=\frac{2p}{\pi}$. This has a unique solution $\xi^\phi\in (0,\pi/2)$, namely $$\xi^\phi(p)=\arctan\big(\frac{2p}{\pi}\big)\nearrow\frac{\pi}{2}\quad\text{as $p\nearrow\infty$}$$ $\phi''_p(x)=-\phi_p(x)=0$ on $[0,\pi]$ iff $\tan(x)=-\frac{\pi}{2p}$. This has a unique solution $\xi_\phi\in(\pi/2,\pi)$, namely $$\zeta_\phi(p)=\pi-\arctan\big(\frac{\pi}{2p}\big)\nearrow\pi\quad\text{as $p\nearrow\infty$}$$ Furthermore, $\phi_p$ attains its maximum in $[0,\pi]$ at $\xi^\phi(p)$, and it has a unique zero and inflection point at $\zeta_\phi(p)$.
The function $f_p$ is strictly decreasing on $[0,\pi]$, and has a unique zero on $[0,\pi$, namely $$\zeta_f(p)=2^{-1/p}\pi\nearrow\pi\quad\text{as $p\nearrow\infty$}$$ For $0<p\leq1$, $\zeta_f(p)\in(0,\pi/2]$ whereas for $p>1$, $\zeta_f(p)\in(\pi/2,\pi)$.
Since $f''(x)=-\frac{2p(p-1)}{\pi^2}(x/\pi)^{p-2}$ for $p\neq1$, $f$ is concave if $p>1$ and convex if $p<1$.
For $0<p\leq1$, $g=\phi-f$ is concave and $g(0)=g(1)$, monotone increasing near $0$ and monotone decreasing near $\pi=0$. Thus there cannot be solutions to $g=0$ inside $(0,\pi)$ and so $g\geq0$.
The case $p>1$ involves more work. The following seems to hold $$\frac{\pi}{2}<\zeta_f(p)<\zeta_\phi(p)$$ The left-hand-side is obvious; the second should follow by showing that $$G(p):=\arctan\big(\frac{\pi}{2p}\big)+2^{-1/p}\pi<\pi$$ Clearly $G(p)\xrightarrow{p\rightarrow\infty}\pi$. Taking derivatives on the left-hand-side, for example, yields $$ G'(p)=-\frac{2\pi}{\pi^2+4p^2}+\frac{\pi 2^{-1/p}\log 2}{p^2} $$ I have not checked this carefully, but it seems that indeed $G'(p)>0$. Putting all this together will imply that $g_p>f_p$ on $[0,\zeta\phi(p)]$.
For the interval $[\zeta_\phi(p),\pi]$ we notice that $\phi'_p(\pi)=-\frac{2p}{\pi}=f'(\pi)$. Since $\phi_p$ is convex on this interval and $f_p$ is concave, it must be the $g_p>f_p$ on $[\zeta_\phi(p),\pi)$. This would be the end of the proof.
There might be more clever ways to prove this inequality and I am looking forward to see other attempts.