Covmodel arg bugfix: Always convert nugget to float

[MuellerSeb]

This line of code generated an Error with gstools==1.3.0rc2:

import gstools as gs
print(gs.Matern(nugget=0))

ValueError: Precision not allowed in integer format specifier

Reason: nugget was never converted to float internally and is stored as int (which is totally fine for all calculations), but the printing routine tries to format it as float.

This PR solves this problem and adds some minor additions:

• rescale was not setable after the model was instantiated: added rescale.setter
• rescale didn't show up in the repr string: fixed if rescale is not the default value
• anis and angles where always showing up also for isotropic and unrotated models: fixed

This means, the repr is the exact init sequence to create a model:

from gstools import Matern
exec(f"m={Matern()}")
print(m)

Matern(dim=3, var=1.0, len_scale=1.0, nugget=0.0, nu=1.0)

The conditional repr looks like the following:

from gstools import Matern
print(Matern(rescale=2))
print(Matern(anis=2))
print(Matern(angles=1))
print(Matern(rescale=2, anis=2, angles=1))

Matern(dim=3, var=1.0, len_scale=1.0, nugget=0.0, rescale=2.0, nu=1.0)
Matern(dim=3, var=1.0, len_scale=1.0, nugget=0.0, anis=[1.0, 2.0], nu=1.0)
Matern(dim=3, var=1.0, len_scale=1.0, nugget=0.0, angles=[1.0, 0.0, 0.0], nu=1.0)
Matern(dim=3, var=1.0, len_scale=1.0, nugget=0.0, anis=[1.0, 2.0], angles=[1.0, 0.0, 0.0], rescale=2.0, nu=1.0)

[LSchueler]

Funny little bug :-)
Thanks for taking a look it.
Check the f-strings and then you can happily merge!