Hard bounds

[majafranz]

This PR adresses that specifying bounds for the FOURIER coefficient does not really make sense.
Instead, now, when applying bounds with the FOURIER strategy, hard limits for beta and gamma parameters are set after applying the sine and cosine transformations.
Resulting parameters are in $\beta \in [-\pi / 2, \pi / 2]^p$ and $\gamma \in [0, \pi ]^p$.
If a $\gamma$ parameters is negative, we account for the QAOA symmetries, by flipping the sign of the respective gamma and beta parameter.

Example:
If the $v$ and $u$ parameters result in a gamma parameter of $\gamma_1=-5$, and a beta parameter of $\beta_1=1$ applying the bound results in $\gamma_{1,b}=\pi$ and $\beta_{1, b} = -1$, through flipping and applying bounds.

Additionally, parameters are also normalised according to bounds and symmetries, when computing the annealing schedules from the parameters.

[stroblme]

thanks for the great work! I left a comment but nothing severe.
Also I'll add a commit regarding the default values for the cost fkt.
Because it's an inner fct, I don't think that we need default values here?