[BUG] Fix Levy distribution to conform with scipy interface

[direkkakkar319-ops]

Description:
Fixes #879

What was broken:
Levy.__init__ manually wrote to self.__dict__["loc"] and self.__dict__["scale"]
after the property setters already ran. This caused split-state — after set_params(),
the property returned the updated value but __dict__ retained the stale constructor
value. sklearn.clone() silently returned the wrong parameters.

What this PR does:

• Removes the __dict__ manipulation entirely
• Removes the unnecessary @property / setter boilerplate for loc and scale
• Renames loc → mu to avoid conflict with BaseDistribution's reserved loc property
• Follows the same pattern as Normal and Laplace in the codebase

Tests:

• 65 Levy-specific tests: all passed
• 89 full distribution tests: all passed

[fkiraly]

There were a lot of spurious methods. I removed them, the distribution is hopefully fixed now.

[direkkakkar319-ops]

Hi @fkiraly! Thank you for jumping in and cleaning that up .
Glad the bug was worth fixing. I've learned a lot about how BaseDistribution handles parameters from this.

[direkkakkar319-ops]

Is there anything else you'd like me to address before this PR is ready . Happy to update tests, docs, or anything else needed!

[direkkakkar319-ops]

Hi @fkiraly! I noticed after your cleanup that the base class was still BaseDistribution — changed it to _ScipyAdapter so that _get_scipy_object and _get_scipy_param work correctly.

All 65 Levy tests passing now

[direkkakkar319-ops]

Changed this

Examples -------- >>> from skpro.distributions.levy import Levy >>> dist = Levy(mu=0.0, scale=1.0) >>> dist.mean()

to this

Examples -------- >>> from skpro.distributions.levy import Levy >>> dist = Levy(mu=0.0, scale=1.0) >>> dist.mean() inf

in this file skpro/distributions/levy.py

[direkkakkar319-ops]

Hi @fkiraly soon will fix to pass the test case. Inform you when is done .

[fkiraly]

Finished it.