DeadZoneRegressor support for "constant" effect

[FBruzzesi]

Description

The PR solves #587 by adding support for effect="constant".

Additionally, as discussed in the issue, it removes the need of autograd by implementing derivative by hand (😂) and by using scipy.optimize.minimize for optimization.

This leads to a few potential breaking changing in the API as the following argument are no longer needed:

• n_iter
• stepsize
• check_grad

and the following attributes are no longer available:

• loss_log_
• self.wts_log_
• deriv_log_

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)

Checklist:

• My code follows the style guidelines (flake8)
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation (also to the readme.md)
• I have added tests that prove my fix is effective or that my feature works
• I have added tests to check whether the new feature adheres to the sklearn convention
• New and existing unit tests pass locally with my changes

If you feel your PR is ready for a review, ping @koaning or @MBrouns.

[FBruzzesi]

Constant effect gets not even close to target values 😞

[koaning]

LGTM. Nice to have autograd out of here.