Fix python vectorizer, prompt and other little things

[jfrverdasca]

What this PR change:

• Add file path to embeddings to allow retrieve files by its name in prompt (like "Add feature X to this_file.py")
• Using python vectorizer, files with syntax errors not are loaded into embeddings in chunks.
• Removed useless information added to embeddings when using python vectorizer (due to the way embeddings are used now)
• Add line numbers to files sent in context following almost the same format of the console command nl -ba file.txt
• Changed prompt to a more complete one adding a little project overview, general guidance, task guidelines, expected code style, testing and resources description showing better LLM results.
• Sort files and steps: Now doesn't matter how LLM results response are, they are always sorted by file name and reverse line order to avoid operations that can cause line shifting
• Added file_handler.py functions tests to ensure that file handle functions act as expected

Reviewers:

• @kenvontucky
• @JdFSilva
• @cld-vasconcelos

Issues:

• Ensure valid syntax before pull request #178 (most of the changed don't meet what is described in the issue)

Print screens showing the changes:

Findings:

Using Chunck Vectorizer and GPT embeddings model (described in the image) this is the result of retrieved embeddings for prompt Add created_at and updated_at fields to User model:

There are the results for the same prompt using nomicembeddings model (described in the image):

As we can see, the models.py file isn't retrieved and the other files are just garbage for what was asked