better way to initialize LinearRegression with params

[mathause]

Creating a new LinearRegression instance is a bit awkward as the params have to be assigned in a separate call.

lr = LinearRegression()
lr.params = params

We should enable a better approach. I see 3 options. I probably have a slight preference for option 2, but no strong opinion.

Option 1

Allow passing params to __init__:

lr = LinearRegression(params)

Option 2

Create a classmethod

lr = LinearRegression.from_params(params)

See

mesmer/mesmer/stats/_linear_regression.py

Lines 157 to 158 in ea98d40

	@classmethod
	def from_netcdf(cls, filename, **kwargs):

Option 3

Split the class into methods...