Add change supporting unit testing

[andystaples]

This PR add support for unit testing Durable client functions, orchestrators, and entities without any change needed to azure-functions-python-library or azure-functions-python-worker.

This is achieved by augmenting the handler returned by the Durable extension into a callable class, with a property storing the original user-defined function.

This way, the worker is still able to process these function types as normal, by passing the context to be parsed by our extension code, but when these new handlers are returned during unit testing, users can access the original function for testing.

Syntactically, here is what that looks like:
Currently, for non-Durable functions and Durable activity functions, even though the Functions decorators replace the original function (see the "Meaning of decorators" section here), users can access the original function either by calling the function directly:

  async def test_function_1(self, client, req):
    function_result = function_1(...)

or by using methods on the wrapped version like so:

  async def test_function_1(self, client, req):
    function_result = function_1.build().get_user_function()(...)

This PR would change the signature of this boiler plate code slightly. We would not preserve the behavior of calling the function directly, and the way to access the original function would look like this:

# Client function test
  async def test_my_client_function(self, client, req):
    function_result = my_client_function.build().get_user_function().client_function(...)

# Orchestrator function test
  async def test_my_orchestrator_function(self, client, req):
    function_result = my_orchestrator_function.build().get_user_function().orchestrator_function(...)

# Entity function test
  async def test_my_entity_function(self, client, req):
    function_result = my_entity_function.build().get_user_function().entity_function(...)

Discussion about whether this solution is adequate and possible improvements is welcome.

[andystaples]

Will fix linting once the implementation is closer to complete

[cgillum]

I'm fine with this approach. Can we add some sample code that shows how to write unit tests (and make sure it works)?

[andystaples]

I have deliberately excluded the tests from running in the GitHub YAML. The reasons for this are numerous, primarily because the tests are run from the project directory of this project, not from the sample app directories, which leads to two problems:

1. Relative directory module imports within the sample projects will not work as intended. This can be fixed with the changes from the commit #4584c90 in this PR, but this is not neccesary when users will interact with the samples and leads to cluttered code
2. The requirements.txt for the Durable project does not contain all the requirements for the test app projects. Primarily, azure-storage-blob is missing. Adding azure-storage-blob to the root-level requirements.txt for this project would likely add an unnecessary package dependency for customers installing this module using pip.

The "correct" way to run the tests from the samples project would be to add a separate GitHub workflow/action to build and test the samples projects independently. I can do this, if there is enough value there, but I have verified that this sample code tests successfully locally when built and run from the sample app directories.

[cgillum]

My prayer is that someday I will understand how module imports in Python work. I get horribly confused every time I run into issues like this.

The "correct" way to run the tests from the samples project would be to add a separate GitHub workflow/action to build and test the samples projects independently. I can do this, if there is enough value there

I think it's worth doing. I sometimes like to think that a feature doesn't actually work unless there is a test proving that it works.

[andystaples]

I think it's worth doing. I sometimes like to think that a feature doesn't actually work unless there is a test proving that it works.

Done!