Avoid duplicated generated code when splitting dataset

[felipealencar]

When a user executes the Split Data feature more than once, the code generation repeats the X.sample and X.drop lines. This behavior might not be desired in the final Python code.

Possible solutions:

1. Implement an application state to check if code was generated with the split method already.
2. Implement a find and update method at the CodeBlock class.
3. Simply clear the CodeBlock state after each code generation.

Example:

[kate-holdener]

Indeed! In fact, for any duplicate "next action", the user chooses, duplicated code will be generated, so I think this issue is more general than the Split Data feature.

I think we either need to filter generated code, to remove duplication OR disallow the user to do duplicated actions (somewhat related to issue #3). If we go with disallowing the user to do duplicated actions, we'd need to add an "Undo" option, to allow the user to go back to previous steps.

[kate-holdener]

Another possible solution: update CodeBlock (in model/code_blocks.py) to have an "id" associated with it. When new block is added by calling add_next_block in AllBlocks, if a block with the same id already exists in AllBlocks, the old version of the block should be removed and the new one added to the end. AllBlocks::add_next_block is called from application/code_generator.py _create_new_block function, which gets called from _parse_and_execute. We can use the 'template' name passed to _parse_and_execute to generate the id: keep a dictionary that maps a template name to a unique id. If the template name is not in the dictionary, add it with a new unique id.