[BUG] [Tech Debt] BFS is slower than we'd like

[bee-san]

#107

Every Text struct contains a field:

text: Vec<String>,

Which contains every decoding (even if it's just 1 decoding).

After each iteration of BFS we loop over all of our vectors we have collected and flatten them from a Vec<Vec<Text>> to a Vec<Text> where text: Vec<String> has a single element (meaning we can index into the 0th element [0] and get the result).

        let mut new_strings_to_be_added = Vec::new();
        for textStruct in new_strings{
            for decoded_text in textStruct.text{
                new_strings_to_be_added.push(
                    Text {
                        text: vec![decoded_text],
                        // quick hack
                        path: textStruct.path.clone(),
                    }
                )
            }

This introduces some issues:

1. It is not efficient to have a O(n^2) loop after our big main loop
2. We are using .clone() here which is slow
3. It is not efficient to make a new vector with just 1 element
4. It does not support nice paths. If it was Caesar, it won't say "Caesar with shift of 13" for example

This is a hack to get us to support this feature.

In the future we might want to look at:

Editing this bit of code so it supports Vectors of Vectors:
https://github.com/bee-san/Ares/blob/1bcc994052db17e7db948055d012810353c76f4d/src/searchers/bfs.rs#LL32-L33C69

Or adding another loop below this code so it turns into a O(n^2) nested loop:

Ares/src/searchers/bfs.rs

Line 26 in 1bcc994

while !current_strings.is_empty() {