Add unit tests for init, transect_analysis, and stratified_summary Survey-class methods

[brandynlucca]

This PR includes new unit tests for the following Survey class methods and related functions:

init

• load_survey_data
• This does not include a standalone test for the validate_data_columns and read_validated_data functions nested within this method
• biometric_distributions

transect_analysis

• fit_binned_length_weight_relationship
• strata_sex_weight_proportions
• strata_age_binned_weight_proportions
• nasc_to_biomass_conversion
• Functions nested within each of these methods have also had new tests written for more specific/granular examples

stratified_summary

• Functions nested within each of these methods have also had new tests written for more specific/granular examples

Other

• Some functions such as those contained within echopop.computation.operations are used across multiple methods
• Tests have been added for each of these (e.g. group_merge, meld, stretch) using both the current monkey-patched implementation as well as the future updated use as 'normal' functions (i.e. dataframe.method( args ) --> method( dataframe , args )
• Additional utility test functions (i.e. dictionary_shape_equal) were added for testing equality in structures between nested dictionaries
• Small tweaks were included to the actual functions to account for some tested edge cases that caused tests to fail, such as unexpected NaN outputs

[brandynlucca]

@leewujung: okay, so after futzing with the library versions, I believe I have squared away the issues I was running into regarding the test failures online (but passing locally).

[leewujung]

I guess this will be something that gets resolved once #215 is merged?

[brandynlucca]

Yep! The plan is to uncomment this once the kriging rework is completed re: #202.

[leewujung]

Hey @brandynlucca : The tests look great! I have only a couple embedded comments, and those below are applicable more generally so I put them in the overall comment here:

There are a few tests you could use the same approach for testing dtype and values separately like in this comment.

Just a note on what we discussed that rtol=1e-1 is a pretty big tolerance, which may signal that something is going on underneath that is not quite "equal", since the determination is based on the following:

For finite values, isclose uses the following equation to test whether two floating point values are equivalent.

absolute(a - b) <= (atol + rtol * absolute(b))

There are a few tests in which similar large rtol is used, like in test_stratified_transect_statistic, so it'd be good to check what's going on.

Lastly, I think you could move the functions in utility_testing_functions.py to conftest.py as fixtures, so that they can be reused in the tests -- see Fixtures are reusable.