Allow LogParabola to be defined using natural log

[HealthyPear]

In some experiments or software the definition of a LogParabola spectral model might brequire to use the natural logarithm instead of the one in base 10.

This PR allows to instantiate such a model with the addition of a new class attribute analog to the one used by gammapy.modeling.models.LogParabolaSpectralModel

[morcuended]

Looks good

[morcuended]

Maybe here say that the default is base 10 log

[maxnoe]

I don't particularly like the keyword name here, why not call this "log_base" or just pass the log function?

E.g. log=np.log10 vs. log=np.log.

[maxnoe]

Ah, I think you took the name from https://docs.gammapy.org/0.19/api/gammapy.modeling.models.LogParabolaSpectralModel.html#gammapy.modeling.models.LogParabolaSpectralModel.from_log10

But this is not a keyword, it's an alternative constructor, that's why it starts with from_.

[HealthyPear]

yes, the idea was to keep a nomenclature analog to gammapy which is what people tends to use more lately

[maxnoe]

But it doesn't make sense to use the name of a class method for the name of a keyword

[maxnoe]

You could implement the same (or rather the inverse) logic as gammapy, e.g. offer a from_ln method that converts the parameters to base 10.

Or you change the name to log, maybe with an enum like this:

class LogType(Enum):
    LOG10 = auto()
    LN = auto()

class LogParabola:
    def __init__(self, ..., log=LogType.LOG10)