Refactoring dataset upload

[lrosemberg]

Description

In PR #266, the way it was implemented changes the way the CLI displays the output, and also causes an error when trying to parse the response from upload_image.

In addition, some methods like project.single_upload had too many responsibilities, which made it difficult to handle exceptions correctly.

That said, what does this PR do?

• Shifts the responsibility for parsing the error correctly to the rfapi layer
• Makes only the rfapi layer throw exceptions
• Creates new methods project.upload_image and project.save_annotation, these start to be used inside project.single_upload and workflow.upload_dataset, isolating a good part of the code and keeping the functions with single responsibility
• Fixes the CLI output
• Fixes the bug introduced in @266 that fails to parse some image upload responses (more information here: fix: raise UploadErrors for images and annotations #266 (comment))
• Creates some test cases to validate project.upload

How has this change been tested, please provide a testcase or example of how you tested the change?

• Created some tests in the code to validate the project.upload
• Downloaded a dataset, edited a annotation to fail, and uploaded using the CLI, results below:

[lrosemberg]

@tonylampada I believe that the best way to solve this is to create exceptions for each type of error, making them be thrown from the rfapi layer, so that project.single_upload "breaks" and does not fail silently.

To be able to capture errors correctly in the CLI, I divided the responsibilities of project.single_upload into project.upload_image and project.save_annotation, so that I can reuse them both within project.single_upload and within workspace.upload_dataset, keeping the CLI outputs the way they were before.

[tonylampada]

FANTASTIC.
Thank you @lrosemberg