ScalePart value checking

[Gautzilla]

We just found a bug with @PaulCarvaillo @PCarvaillo, where computing a spectrogram with a Scale that has incorrect parameters (e.g. a ScalePart with p_max = 100) would lead to a huuuuuuuuge explosion in RAM usage.

This PR prevents the user from instantiating ScalePart instances with incorrect values. The enforced conditions are:

• p_min >= 0
• p_max >= 0
• p_min <= 1
• p_max <= 1
• p_min < p_max
• f_min >= 0
• f_max >= 0
• f_min < f_max

With these checks in ScalePart.__post_init__(), there should be no way of reproducing the RAM explosion experienced before.