[DOC] guidelines for testing stochastic methods

[jpaillard]

Testing the methods implemented in hidimstat is often not straightforward because of their stochastic behavior. It is thus hard to test for equality.
It would be helpful to write a few guidelines/rules in the developer's doc describing "good practices" that will help standardize the test suite.

[lionelkusch]

This was the objective of issue #375 but we don't arrive to any consensus accept not to have unstable tests and to have tests for the coverage.

[bthirion]

Indeed, guidelines are only useful when we have reached a consensus. This is coming, but slowly.