Cache searches

Re-reading a file, generating its tree, then running a query on it is redundant and time consuming. If we can eliminate atleast one of those we could see good performance gains. For example, we could cache filepaths, so we wouldn't need to regenerate their tree again.

Start by tracing where cgrep reads files, generates their trees, and runs queries. Determine which repeated work can be cached without changing search results, then compare performance before and after to verify that redundant processing is reduced.

Written by the indexing model from the issue text.