Extract and process code snippets from user input

[ptelang]

In addition to natural language text, a user query can contain code snippets. Currently, codegate computes vector embeddings for the entire user query including the code snippets. This can lead to poor similarity search results.

So codegate should be modified to extract the code snippets from the user query. Bad packages if any should be extracted from the code snippets using tree-sitter (similar to the output code snippets). The natural text portion of the user query should be embedded and used for similarity search.

Related to #475