Improve Linear Regression widget

#5722 suggested additional features for scatter plot. They belong to a separate widget, possibly Linear Regression, which could have a plot of single variable vs. the dependent variable.

This requires some planning. To observe regression lines for individual features, we may also want to add univariate regression (as "naive linear regression" that would assume mutual independence of variables).

See also #5845. We need to decide what kinds of statistics to include. As I argued there, Orange is a tool for generating hypothesis from data and such hypothesis must not be tested on this same data; the least we can do is to not help the users doing it. We should however consider adding Bayesian stuff.

Include also #5789 and #5722.

Start by reviewing the related discussions in #5722, #5845, and #5789. Clarify the scope of a separate Linear Regression widget, including whether it should show univariate regression and which statistics or Bayesian features belong. Done means the widget requirements and statistical boundaries are agreed before implementation.

Written by the indexing model from the issue text.